ciWithControlVariates.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparison with CVX"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "code_folding": [
     2,
     70
    ],
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.7845508143539998,\n",
       " [(550, 0, 0, array([ 0., -1.])),\n",
       "  (520, 0, 0, array([-1.,  0.])),\n",
       "  (91, 2, 0, array([1., 0.])),\n",
       "  (89, 2, 0, array([0., 1.])),\n",
       "  (356, 2, 1, array([0., 1.])),\n",
       "  (394, 2, 1, array([1., 0.]))],\n",
       " ((0.7085859835146853,\n",
       "   {'likelihood': -1.6126430692087834,\n",
       "    'mlelikelihood': -1.6116817858889367,\n",
       "    'Delta': -3225.286629224291,\n",
       "    'sumofone': 0.9999999930710612,\n",
       "    'sumofw': 0.9999999962603696,\n",
       "    'sumofcvs': array([5.94311589e-12, 3.18336545e-09])}),\n",
       "  (0.8438965210521095,\n",
       "   {'likelihood': -1.6126426848635855,\n",
       "    'mlelikelihood': -1.6116817858889367,\n",
       "    'Delta': -3225.286629224291,\n",
       "    'sumofone': 0.9999999851663477,\n",
       "    'sumofw': 1.000000002281821,\n",
       "    'sumofcvs': array([3.59319693e-09, 1.35222765e-08])})),\n",
       " ((0.7085892753962597,\n",
       "   {'betastar': array(21.03340321),\n",
       "    'gammastar': array(-0.00409856),\n",
       "    'deltastar': array([-0.0169518, -0.0169518]),\n",
       "    'kappastar': 0.010868946963806018,\n",
       "    'likelihood': -1.612643313357236,\n",
       "    'mlelikelihood': -1.6116817858889367,\n",
       "    'Delta': -1.6126433146121455}),\n",
       "  (0.8438918800391682,\n",
       "   {'betastar': array(12.15003587),\n",
       "    'gammastar': array(0.66413235),\n",
       "    'deltastar': array([0.32404166, 0.32404166]),\n",
       "    'kappastar': 0.005985138168210654,\n",
       "    'likelihood': -1.6126433152445778,\n",
       "    'mlelikelihood': -1.6116817858889367,\n",
       "    'Delta': -1.6126433146121455})),\n",
       " ((0.7085892471301007, 0.8438918731934413),\n",
       "  ({'gammastar': 21.04989541691046,\n",
       "    'betastar': -0.021049741775770688,\n",
       "    'deltastar': array([-1.43784925e-07, -1.43784925e-07]),\n",
       "    'kappastar': 0.010868717461132394,\n",
       "    'qfunc': <function MLE.MLE.asymptoticconfidenceintervalwithcv.asymptoticconfidenceintervalwithcv.<locals>.<lambda>(c, w, r, cvs)>},\n",
       "   {'gammastar': 11.807173221686057,\n",
       "    'betastar': 1.0073296715912763,\n",
       "    'deltastar': array([-0.01915599, -0.01915599]),\n",
       "    'kappastar': 0.005985305510041945,\n",
       "    'qfunc': <function MLE.MLE.asymptoticconfidenceintervalwithcv.asymptoticconfidenceintervalwithcv.<locals>.<lambda>(c, w, r, cvs)>})))"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class CVXCITest:\n",
    "    @staticmethod\n",
    "    def cvxdualestimate(data, rangefn, sign=1, rmin=0, rmax=1, alpha=0.05):\n",
    "        import cvxpy as cp\n",
    "        import numpy as np\n",
    "        from collections import defaultdict\n",
    "        from scipy.stats import f\n",
    "        from scipy.special import xlogy\n",
    "        \n",
    "        cdict = defaultdict(int)\n",
    "        n = 0\n",
    "        clogc = 0\n",
    "        for (ci, wi, ri, cvsi) in data:\n",
    "            if ci > 0:\n",
    "                cdict[(wi, ri, tuple(cvsi))] += ci\n",
    "            n += ci\n",
    "            clogc += xlogy(ci,ci)\n",
    "            ncvs = len(cvsi)\n",
    "        assert n >= 1\n",
    "        cdict.default_factory = None\n",
    "        \n",
    "        import MLE.MLE\n",
    "        mle = MLE.MLE.estimatewithcv(datagen=lambda: data, rangefn=rangefn, rmin=rmin, rmax=rmax)\n",
    "        mlelikelihood = sum(xlogy(c, mle[1]['qfunc'](c, w, r, cvs)) for c, w, r, cvs in data)\n",
    "        Delta = mlelikelihood - 0.5 * f.isf(q=alpha, dfn=1, dfd=n-1)\n",
    "        \n",
    "        C = np.array([ value for key, value in cdict.items() ])\n",
    "        wvec = np.array([ w for (w, _, _), _ in cdict.items() ])\n",
    "        wrvec = np.array([ w * r for (w, r, _), _ in cdict.items() ])\n",
    "        cvsmat = np.array([ cvs for (_, _, cvs), _ in cdict.items() ])\n",
    "\n",
    "        beta = cp.Variable()\n",
    "        gamma = cp.Variable()\n",
    "        delta = cp.Variable(ncvs)\n",
    "                \n",
    "        objective = (\n",
    "            n * cp.exp((Delta - clogc) / n) * cp.geo_mean(\n",
    "                x=beta + gamma * wvec + cp.matmul(cvsmat, delta) + sign * wrvec,\n",
    "                p=C.astype('int').flatten(order='F'),\n",
    "                max_denom=1<<24\n",
    "            )\n",
    "            -beta\n",
    "            -gamma\n",
    "        )\n",
    "        constraints = [\n",
    "           beta + gamma * w + cp.matmul(np.array(cv), delta) + sign * w * r >= 0\n",
    "           for w, cv in rangefn()\n",
    "           for r in (rmin, rmax)\n",
    "        ]\n",
    "        \n",
    "        prob = cp.Problem(cp.Maximize(objective), constraints)\n",
    "        prob.solve(solver='ECOS', max_iters=200)\n",
    "        \n",
    "        vlb = sign * prob.value            \n",
    "        kappastar = (prob.value + beta.value + gamma.value) / n\n",
    "\n",
    "        Q = np.array([ kappastar * c / (beta.value + gamma.value * w + delta.value.dot(np.array(cvs)) + sign * w * r)\n",
    "                       for (w, r, cvs), c in cdict.items() ])\n",
    "                \n",
    "        return vlb, {\n",
    "            'betastar': beta.value,\n",
    "            'gammastar': gamma.value,\n",
    "            'deltastar': delta.value,\n",
    "            'kappastar': kappastar,\n",
    "            'likelihood': sum(xlogy(C, Q)) / n,\n",
    "            'mlelikelihood': mlelikelihood / n,\n",
    "            'Delta': Delta / n,\n",
    "        }\n",
    "        \n",
    "    @staticmethod\n",
    "    def cvxestimate(data, rangefn, sign=1, rmin=0, rmax=1, alpha=0.05):\n",
    "        import cvxpy as cp\n",
    "        import numpy as np\n",
    "        from collections import defaultdict\n",
    "        from scipy.stats import f\n",
    "        from scipy.special import xlogy\n",
    "            \n",
    "        cdict = defaultdict(int)\n",
    "        n = 0\n",
    "        for (ci, wi, ri, cvsi) in data:\n",
    "            if ci > 0:\n",
    "                cdict[(wi, ri, tuple(cvsi))] += ci\n",
    "            n += ci\n",
    "        assert n >= 1\n",
    "        for w, cv in rangefn():\n",
    "            for r in (rmin, rmax):\n",
    "                cdict[(w, r, tuple(cv))] += 0\n",
    "        cdict.default_factory = None\n",
    "        \n",
    "        import MLE.MLE\n",
    "        mle = MLE.MLE.estimatewithcv(datagen=lambda: data, rangefn=rangefn, rmin=rmin, rmax=rmax)\n",
    "        mlelikelihood = sum(xlogy(c, mle[1]['qfunc'](c, w, r, cvs)) for c, w, r, cvs in data)\n",
    "        Delta = mlelikelihood - 0.5 * f.isf(q=alpha, dfn=1, dfd=n-1)\n",
    "                       \n",
    "        C = np.array([ value / n for key, value in cdict.items() ])\n",
    "        wvec = np.array([ w for (w, _, _), _ in cdict.items() ])\n",
    "        wrvec = np.array([ w * r for (w, r, _), _ in cdict.items() ])\n",
    "        cvsmat = np.array([ cvs for (_, _, cvs), _ in cdict.items() ])\n",
    "        Q = cp.Variable(len(C))\n",
    "        \n",
    "        prob = cp.Problem(cp.Minimize(sign*cp.sum(cp.multiply(wrvec, Q))), [\n",
    "            cp.sum(cp.multiply(wvec, Q)) == 1,\n",
    "            cp.sum(Q) == 1,\n",
    "            cp.matmul(cvsmat.T, Q) == 0,\n",
    "            cp.sum(cp.multiply(C, cp.log(Q))) >= Delta / n \n",
    "        ])\n",
    "        prob.solve(solver='ECOS')\n",
    "        vbound = sign*prob.value\n",
    "            \n",
    "        return vbound, { \n",
    "#             'qstar': { key: q for (key, c), q in zip(cdict.items(), Q.value) if c > 0 },\n",
    "#             'qex': { key: q for (key, c), q in zip(cdict.items(), Q.value) if c == 0 and q > 0},\n",
    "            'likelihood': np.sum(xlogy(C, Q.value)),\n",
    "            'mlelikelihood': mlelikelihood / n,\n",
    "            'Delta': Delta,\n",
    "            'sumofone': np.sum(Q.value),\n",
    "            'sumofw': wvec.dot(Q.value),\n",
    "            'sumofcvs': np.matmul(cvsmat.T, Q.value)\n",
    "        }\n",
    "    \n",
    "def flass():\n",
    "    from importlib import reload\n",
    "    import numpy as np\n",
    "    import environments.DoubleDouble\n",
    "    import MLE.MLE\n",
    "    reload(MLE.MLE)\n",
    "    \n",
    "    ddm = environments.DoubleDouble.DoubleDouble(numactions=2, seed=45, wsupport=[0, 2, 1000], expwsq=100)\n",
    "    (truevalue, data) = ddm.samplewithcvs(2000)\n",
    "    return (truevalue, data, \n",
    "            ( \n",
    "                CVXCITest.cvxestimate(data, sign=1, rangefn=ddm.range),\n",
    "                CVXCITest.cvxestimate(data, sign=-1, rangefn=ddm.range),\n",
    "            ),\n",
    "            (\n",
    "                CVXCITest.cvxdualestimate(data, sign=1, rangefn=ddm.range),\n",
    "                CVXCITest.cvxdualestimate(data, sign=-1, rangefn=ddm.range),\n",
    "            ),  \n",
    "            MLE.MLE.asymptoticconfidenceintervalwithcv(lambda: data, rangefn=ddm.range, raiseonerr=True)\n",
    "           ) \n",
    "\n",
    "flass()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "code_folding": [
     0
    ]
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 14/14 [02:24<00:00, 21.14s/it]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'badcvx': 0, 'total': 154}"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def testasymptoticci():\n",
    "    from importlib import reload\n",
    "    from math import ceil\n",
    "    import environments.DoubleDouble \n",
    "    import MLE.MLE\n",
    "    import numpy as np\n",
    "    from tqdm import tqdm as tqdm\n",
    "    import sys\n",
    "    from pprint import pformat\n",
    "    from scipy.stats import f\n",
    "    from scipy.special import xlogy\n",
    "    \n",
    "    reload(environments.DoubleDouble)\n",
    "    reload(MLE.MLE)\n",
    "    \n",
    "    wsupport = [ 0, 2, 10 ]\n",
    "    wmin, wmax = wsupport[0], wsupport[-1]\n",
    "    env = environments.DoubleDouble.DoubleDouble(numactions=3, seed=45, wsupport=wsupport, expwsq=wmax/2)\n",
    "    \n",
    "    total = 0\n",
    "    badcvx = 0\n",
    "\n",
    "    for ndata in tqdm(map(ceil, np.logspace(1, 7, 14)), file=sys.stderr, total=14):\n",
    "        for i in range(11):\n",
    "            (truevalue, data) = env.samplewithcvs(ndata)\n",
    "             \n",
    "            (lb, ub), (qlb, qub) = MLE.MLE.asymptoticconfidenceintervalwithcv(lambda: data, rangefn=env.range, \n",
    "                                                                              raiseonerr=True)            \n",
    "            \n",
    "            \n",
    "            assert lb >= 0 and ub <= 1 and lb <= ub, pformat({(lb, ub)})\n",
    "            \n",
    "            total += 1\n",
    "            try:\n",
    "                cvxlb, cvxqlb = CVXCITest.cvxdualestimate(data, sign=1, rangefn=env.range)\n",
    "                cvxub, cvxqub = CVXCITest.cvxdualestimate(data, sign=-1, rangefn=env.range)\n",
    "            except:\n",
    "                badcvx += 1\n",
    "                continue\n",
    "\n",
    "            assert (np.allclose(cvxlb, lb, atol=5e-3) \n",
    "                    and np.allclose(cvxub, ub, atol=5e-3)\n",
    "                   ), pformat(\n",
    "            {\n",
    "                'data': data,\n",
    "                'wmin': wmin,\n",
    "                'wmax': wmax,\n",
    "                'cvxlb': cvxlb,\n",
    "                'cvxub': cvxub,\n",
    "                'cvxqlb': cvxqlb,\n",
    "                'cvxqub': cvxqub,\n",
    "                'lb': lb,\n",
    "                'ub': ub,\n",
    "                'qlb': qlb,\n",
    "                'qub': qub,\n",
    "            })\n",
    "    \n",
    "    return { 'badcvx': badcvx, 'total': total }\n",
    "    \n",
    "testasymptoticci()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparison with Baselines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "code_folding": [
     6,
     22,
     40
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "****** SNIPS Bootstrap ******\n",
      "(10, {'cov': 0.61, 'covstd': 0.0493, 'ubcov': 0.75, 'lbcov': 0.86, 'width': 0.4722, 'widthstd': 0.0393, 'widthlo': 0.0, 'widthhi': 1.0, 'ub': 0.6923, 'lb': 0.22})\n",
      "(19, {'cov': 0.64, 'covstd': 0.0485, 'ubcov': 0.85, 'lbcov': 0.79, 'width': 0.419, 'widthstd': 0.0319, 'widthlo': 0.0, 'widthhi': 0.8417, 'ub': 0.744, 'lb': 0.3251})\n",
      "(34, {'cov': 0.77, 'covstd': 0.0425, 'ubcov': 0.94, 'lbcov': 0.83, 'width': 0.3369, 'widthstd': 0.0198, 'widthlo': 0.0, 'widthhi': 0.5973, 'ub': 0.7229, 'lb': 0.386})\n",
      "(62, {'cov': 0.77, 'covstd': 0.0425, 'ubcov': 0.88, 'lbcov': 0.89, 'width': 0.2584, 'widthstd': 0.0137, 'widthlo': 0.0, 'widthhi': 0.3933, 'ub': 0.6402, 'lb': 0.3818})\n",
      "(113, {'cov': 0.83, 'covstd': 0.0379, 'ubcov': 0.91, 'lbcov': 0.92, 'width': 0.2189, 'widthstd': 0.0106, 'widthlo': 0.0612, 'widthhi': 0.3015, 'ub': 0.6111, 'lb': 0.3922})\n",
      "(207, {'cov': 0.8, 'covstd': 0.0404, 'ubcov': 0.92, 'lbcov': 0.88, 'width': 0.1687, 'widthstd': 0.0113, 'widthlo': 0.0369, 'widthhi': 0.219, 'ub': 0.6211, 'lb': 0.4524})\n",
      "(380, {'cov': 0.78, 'covstd': 0.0418, 'ubcov': 0.86, 'lbcov': 0.92, 'width': 0.1275, 'widthstd': 0.0092, 'widthlo': 0.0214, 'widthhi': 0.1599, 'ub': 0.5064, 'lb': 0.379})\n",
      "(696, {'cov': 0.6, 'covstd': 0.0495, 'ubcov': 0.82, 'lbcov': 0.78, 'width': 0.0944, 'widthstd': 0.0077, 'widthlo': 0.0201, 'widthhi': 0.1153, 'ub': 0.5473, 'lb': 0.4529})\n",
      "(1275, {'cov': 0.55, 'covstd': 0.0503, 'ubcov': 0.8, 'lbcov': 0.75, 'width': 0.1014, 'widthstd': 0.0145, 'widthlo': 0.0144, 'widthhi': 0.4837, 'ub': 0.5911, 'lb': 0.4897})\n",
      "(2336, {'cov': 0.57, 'covstd': 0.05, 'ubcov': 0.84, 'lbcov': 0.73, 'width': 0.1179, 'widthstd': 0.0161, 'widthlo': 0.0136, 'widthhi': 0.55, 'ub': 0.5655, 'lb': 0.4476})\n",
      "(4282, {'cov': 0.43, 'covstd': 0.05, 'ubcov': 0.74, 'lbcov': 0.69, 'width': 0.0861, 'widthstd': 0.0119, 'widthlo': 0.0118, 'widthhi': 0.4187, 'ub': 0.5981, 'lb': 0.512})\n",
      "(7848, {'cov': 0.53, 'covstd': 0.0504, 'ubcov': 0.79, 'lbcov': 0.74, 'width': 0.1226, 'widthstd': 0.012, 'widthlo': 0.0076, 'widthhi': 0.3123, 'ub': 0.5726, 'lb': 0.45})\n",
      "(14385, {'cov': 0.63, 'covstd': 0.0488, 'ubcov': 0.86, 'lbcov': 0.77, 'width': 0.1162, 'widthstd': 0.0088, 'widthlo': 0.009, 'widthhi': 0.2464, 'ub': 0.5902, 'lb': 0.474})\n",
      "(26367, {'cov': 0.72, 'covstd': 0.0454, 'ubcov': 0.84, 'lbcov': 0.88, 'width': 0.0826, 'widthstd': 0.006, 'widthlo': 0.0053, 'widthhi': 0.1707, 'ub': 0.5172, 'lb': 0.4347})\n",
      "(48330, {'cov': 0.87, 'covstd': 0.034, 'ubcov': 0.95, 'lbcov': 0.92, 'width': 0.0802, 'widthstd': 0.0034, 'widthlo': 0.0081, 'widthhi': 0.1261, 'ub': 0.5725, 'lb': 0.4923})\n",
      "(88587, {'cov': 0.88, 'covstd': 0.0328, 'ubcov': 0.94, 'lbcov': 0.94, 'width': 0.0609, 'widthstd': 0.0024, 'widthlo': 0.0113, 'widthhi': 0.1052, 'ub': 0.5, 'lb': 0.439})\n",
      "(162378, {'cov': 0.93, 'covstd': 0.0258, 'ubcov': 0.97, 'lbcov': 0.96, 'width': 0.0456, 'widthstd': 0.0016, 'widthlo': 0.0183, 'widthhi': 0.0689, 'ub': 0.52, 'lb': 0.4744})\n",
      "(297636, {'cov': 0.91, 'covstd': 0.0289, 'ubcov': 0.97, 'lbcov': 0.94, 'width': 0.0364, 'widthstd': 0.0012, 'widthlo': 0.0164, 'widthhi': 0.0541, 'ub': 0.5302, 'lb': 0.4938})\n",
      "(545560, {'cov': 0.95, 'covstd': 0.022, 'ubcov': 0.97, 'lbcov': 0.98, 'width': 0.0267, 'widthstd': 0.0007, 'widthlo': 0.0129, 'widthhi': 0.0376, 'ub': 0.4937, 'lb': 0.467})\n",
      "(1000000, {'cov': 0.96, 'covstd': 0.0198, 'ubcov': 0.99, 'lbcov': 0.97, 'width': 0.0203, 'widthstd': 0.0006, 'widthlo': 0.0108, 'widthhi': 0.0287, 'ub': 0.4794, 'lb': 0.4591})\n"
     ]
    }
   ],
   "source": [
    "from importlib import reload\n",
    "import MLE.MLE\n",
    "\n",
    "reload(MLE.MLE)\n",
    "\n",
    "class CITest:\n",
    "    class ClopperPearson:\n",
    "        @staticmethod\n",
    "        def interval(data, wmax, alpha, **kwargs):    \n",
    "            from scipy.special import betaincinv\n",
    "\n",
    "            k = sum((c*w/wmax)*r for c, w, r in data)\n",
    "            n = sum(c for c, w, _ in data)\n",
    "\n",
    "            lb = betaincinv(k, n - k + 1, alpha/2) if k > 0 and n - k + 1 > 0 else 0\n",
    "            ub = betaincinv(k + 1, n - k, 1-alpha/2) if n > k else 1\n",
    "            \n",
    "            lb = min(1, max(0, wmax * lb))\n",
    "            ub = min(1, max(0, wmax * ub))\n",
    "\n",
    "            return min(lb, ub), max(lb, ub)\n",
    "        \n",
    "    class AsymptoticGaussian:\n",
    "        def interval(data, wmax, alpha, **kwargs):\n",
    "            from math import sqrt\n",
    "            from scipy.stats import norm  \n",
    "            \n",
    "            scale = norm.ppf(1-alpha/2)\n",
    " \n",
    "            n = sum(c for c, _, _ in data)\n",
    "            tot = sum(c*w*r for c, w, r in data)\n",
    "            muhat = tot / n\n",
    "            totsq = sum(c*(w*r - muhat)**2 for c, w, r in data)\n",
    "            sigmahat = sqrt(totsq) / n\n",
    "            \n",
    "            lb = min(1, max(0, muhat - scale*sigmahat))\n",
    "            ub = min(1, max(0, muhat + scale*sigmahat))\n",
    "            \n",
    "            return min(lb, ub), max(lb, ub)\n",
    "\n",
    "class SNIPS:\n",
    "    @staticmethod\n",
    "    def estimate(data, **kwargs):\n",
    "        effn = sum(c*w for c, w, _ in data)\n",
    "        return 0.5 if effn == 0 else sum(c*w*r for c, w, r in data) / effn\n",
    "        \n",
    "def produceresults(env, method, minexp=1, maxexp=6, numpts=15, ndataperpt=100):\n",
    "    from math import ceil\n",
    "    import numpy\n",
    "    \n",
    "    rangefn = env.range\n",
    "\n",
    "    for ndata in map(ceil, numpy.logspace(minexp, maxexp, numpts)):\n",
    "        bounds=[]\n",
    "        for i in range(1, ndataperpt+1):\n",
    "            (truevalue, data) = env.samplewithcvs(ndata)\n",
    "\n",
    "            ci = None\n",
    "            try:\n",
    "                ci = method(data=data, rangefn=rangefn, alpha=0.05)\n",
    "                assert numpy.isfinite(ci[0]) and numpy.isfinite(ci[1])\n",
    "                assert numpy.isfinite(ci[0]) and numpy.isfinite(ci[1])\n",
    "                assert ci[1] >= ci[0]\n",
    "                assert ci[0] >= -1e-4\n",
    "                assert ci[1] <= 1 + 1e-4\n",
    "#                 assert ci[1] >= truevalue\n",
    "#                 assert ci[0] <= truevalue\n",
    "            except:\n",
    "                print('truevalue was {}'.format(truevalue))\n",
    "                print('data was {}'.format(data))\n",
    "                print('ci was {}'.format(ci))\n",
    "                raise\n",
    "                                                \n",
    "            bounds.append(\n",
    "                ( truevalue,\n",
    "                  data,\n",
    "                  [ \n",
    "                      ci[0], 'n/a', ci[1]\n",
    "                  ]\n",
    "                )  \n",
    "            )\n",
    "            \n",
    "        uppers = [ d[2][2] for d in bounds ]\n",
    "        lowers = [ d[2][0] for d in bounds ]\n",
    "        upperbounded = [ 1 if d[0] <= d[2][2] else 0 for d in bounds ]\n",
    "        lowerbounded = [ 1 if d[2][0] <= d[0] else 0 for d in bounds ]\n",
    "        covered = [ 1 if u*l > 0 else 0 for (u, l) in zip(upperbounded, lowerbounded) ]\n",
    "        width = [ d[2][2] - d[2][0] for d in bounds ]    \n",
    "        yield (ndata,\n",
    "                { \n",
    "                    'cov': numpy.mean(covered),\n",
    "                    'covstd': numpy.std(covered, ddof=1) / numpy.sqrt(len(covered) - 1),\n",
    "                    'ubcov': numpy.mean(upperbounded),\n",
    "                    'lbcov': numpy.mean(lowerbounded),\n",
    "                    'width': numpy.mean(width),\n",
    "                    'widthstd': numpy.std(width, ddof=1) / numpy.sqrt(len(width) - 1),\n",
    "                    'widthlo': numpy.quantile(width, q=[0.05])[0],\n",
    "                    'widthhi': numpy.quantile(width, q=[0.95])[0],\n",
    "                    'ub': numpy.mean(uppers),\n",
    "                    'lb': numpy.mean(lowers),\n",
    "                },\n",
    "              )\n",
    "\n",
    "def bootstrapci(data, rangefn, alpha, **kwargs):\n",
    "    import numpy \n",
    "    from math import ceil\n",
    "    \n",
    "    what = kwargs.pop('what')\n",
    "    assert what == 'mle' or what == 'mlecv' or what == 'snips'\n",
    "    \n",
    "    def doit(counts, data):\n",
    "        if what == 'mle':\n",
    "            vhat, qstar = MLE.MLE.estimate(lambda: ((c, w, r) for c, (_, w, r, cvs) in zip(counts, data)),\n",
    "                                           wmin=rangefn('wmin'),\n",
    "                                           wmax=rangefn('wmax'),\n",
    "                                           raiseonerr=True)  \n",
    "        elif what == 'snips':\n",
    "            vhat = SNIPS.estimate([ (c, w, r) for c, (_, w, r, cvs) in zip(counts, data) ])\n",
    "            qstar = { 'vmin': vhat, 'vmax': vhat }\n",
    "        else:\n",
    "            vhat, qstar = MLE.MLE.estimatewithcv(lambda: ((c, w, r, cvs) for c, (_, w, r, cvs) in zip(counts, data)),\n",
    "                                                 rangefn=rangefn, \n",
    "                                                 raiseonerr=True)\n",
    "            \n",
    "        return qstar\n",
    "                \n",
    "    p = numpy.array([ c for c, _, _, _ in data ], dtype='float64')\n",
    "    qmu = doit(p, data)\n",
    "    \n",
    "    n = numpy.sum(p)\n",
    "    p /= n\n",
    "    numsamples = int(ceil(32.0 / alpha))\n",
    "    state = numpy.random.RandomState(90210)\n",
    "    resamples = state.multinomial(n=n, pvals=p, size=numsamples)\n",
    "    uppers = []\n",
    "    lowers = []\n",
    "    for counts in resamples:\n",
    "        qstar = doit(counts, data)   \n",
    "        uppers.append(qstar['vmax'] - qmu['vmax'])\n",
    "        lowers.append(qstar['vmin'] - qmu['vmin'])\n",
    "        \n",
    "    ub = qmu['vmax'] + numpy.quantile(uppers, 1 - alpha / 2.0)\n",
    "    lb = qmu['vmin'] + numpy.quantile(lowers, alpha / 2.0)\n",
    "    \n",
    "    return min(1, max(0, lb)), min(1, max(0, ub))\n",
    "        \n",
    "def getenv():\n",
    "    import environments.DoubleDouble \n",
    "    wsupport = [ 0, 2, 1000 ]\n",
    "    env = environments.DoubleDouble.DoubleDouble(numactions=5, seed=45, wsupport=wsupport, expwsq=100)\n",
    "    return env\n",
    "    \n",
    "allres = []\n",
    "for (name, method) in [ \n",
    "                        ('SNIPS Bootstrap', lambda **kwargs: bootstrapci(what='snips', **kwargs)),\n",
    "#                         ('MLE Bootstrap', lambda **kwargs: bootstrapci(what='mle', **kwargs)),\n",
    "#                         ('MLE+CV Bootstrap', lambda **kwargs: bootstrapci(what='mlecv', **kwargs)),                  \n",
    "#                         ('MLE CI CV', lambda data, **kwargs: MLE.MLE.asymptoticconfidenceintervalwithcv(datagen=lambda: data, **kwargs)[0]),\n",
    "#                         ('MLE CI', lambda data, rangefn, **kwargs: MLE.MLE.asymptoticconfidenceinterval(datagen=lambda: [(c,w,r) for c, w, r, _ in data], wmin=rangefn('wmin'), wmax=rangefn('wmax'), **kwargs)[0]),\n",
    "#                         ('Asymptotic Gaussian', lambda data, rangefn, alpha, **kwargs: CITest.AsymptoticGaussian.interval([(c,w,r) for c, w, r, _ in data], rangefn('wmax'), alpha)),\n",
    "#                         ('Binomial CI', lambda data, rangefn, alpha, **kwargs: CITest.ClopperPearson.interval([(c,w,r) for c, w, r, _ in data], rangefn('wmax'), alpha)),\n",
    "                      ]:\n",
    "    print('****** {} ******'.format(name))\n",
    "    res = []\n",
    "    for zzz in produceresults(getenv(), method, ndataperpt=100, numpts=20):\n",
    "        res.append(zzz)\n",
    "        print('{}'.format(( zzz[0], { y: round(z, 4) for y, z in zzz[1].items() })), flush=True)\n",
    "    wmax = getenv().range('wmax')\n",
    "    allres.append((name, [(x[0] / wmax, x[1]) for x in res]))\n",
    "    del wmax\n",
    "    import pickle\n",
    "    pickle.dump( allres, open( \"epsilongreedy_ciwithcv_allres_booty.p\", \"wb\" ) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "code_folding": [
     5,
     40
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "****** MLE Bootstrap ******\n",
      "(10, {'cov': 0.981, 'covstd': 0.0043, 'ubcov': 0.988, 'lbcov': 0.993, 'width': 0.798, 'widthstd': 0.0066, 'widthlo': 0.3994, 'widthhi': 1.0, 'ub': 0.9012, 'lb': 0.1032})\n",
      "(19, {'cov': 0.981, 'covstd': 0.0043, 'ubcov': 0.99, 'lbcov': 0.991, 'width': 0.6547, 'widthstd': 0.0062, 'widthlo': 0.2651, 'widthhi': 0.8946, 'ub': 0.8323, 'lb': 0.1775})\n",
      "(34, {'cov': 0.986, 'covstd': 0.0037, 'ubcov': 0.989, 'lbcov': 0.997, 'width': 0.5379, 'widthstd': 0.0048, 'widthlo': 0.2934, 'widthhi': 0.7645, 'ub': 0.78, 'lb': 0.2422})\n",
      "(62, {'cov': 0.992, 'covstd': 0.0028, 'ubcov': 0.994, 'lbcov': 0.998, 'width': 0.4406, 'widthstd': 0.0038, 'widthlo': 0.2383, 'widthhi': 0.6448, 'ub': 0.7151, 'lb': 0.2744})\n",
      "(113, {'cov': 0.991, 'covstd': 0.003, 'ubcov': 0.998, 'lbcov': 0.993, 'width': 0.3559, 'widthstd': 0.003, 'widthlo': 0.2027, 'widthhi': 0.5039, 'ub': 0.6624, 'lb': 0.3065})\n",
      "(207, {'cov': 0.996, 'covstd': 0.002, 'ubcov': 0.999, 'lbcov': 0.997, 'width': 0.2915, 'widthstd': 0.0022, 'widthlo': 0.1873, 'widthhi': 0.41, 'ub': 0.6372, 'lb': 0.3457})\n",
      "(380, {'cov': 0.992, 'covstd': 0.0028, 'ubcov': 0.994, 'lbcov': 0.998, 'width': 0.2437, 'widthstd': 0.0017, 'widthlo': 0.1572, 'widthhi': 0.3362, 'ub': 0.6288, 'lb': 0.3851})\n",
      "(696, {'cov': 0.995, 'covstd': 0.0022, 'ubcov': 0.996, 'lbcov': 0.999, 'width': 0.2037, 'widthstd': 0.0012, 'widthlo': 0.1399, 'widthhi': 0.2668, 'ub': 0.6125, 'lb': 0.4088})\n",
      "(1275, {'cov': 0.995, 'covstd': 0.0022, 'ubcov': 0.998, 'lbcov': 0.997, 'width': 0.1768, 'widthstd': 0.0009, 'widthlo': 0.1286, 'widthhi': 0.2212, 'ub': 0.6011, 'lb': 0.4243})\n",
      "(2336, {'cov': 0.997, 'covstd': 0.0017, 'ubcov': 0.997, 'lbcov': 1.0, 'width': 0.1561, 'widthstd': 0.0007, 'widthlo': 0.119, 'widthhi': 0.1927, 'ub': 0.5803, 'lb': 0.4242})\n",
      "(4282, {'cov': 0.996, 'covstd': 0.002, 'ubcov': 0.998, 'lbcov': 0.998, 'width': 0.1391, 'widthstd': 0.0005, 'widthlo': 0.1098, 'widthhi': 0.1674, 'ub': 0.5538, 'lb': 0.4147})\n",
      "(7848, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.998, 'lbcov': 1.0, 'width': 0.1274, 'widthstd': 0.0004, 'widthlo': 0.105, 'widthhi': 0.1487, 'ub': 0.5589, 'lb': 0.4314})\n",
      "(14385, {'cov': 0.982, 'covstd': 0.0042, 'ubcov': 0.991, 'lbcov': 0.991, 'width': 0.1147, 'widthstd': 0.0006, 'widthlo': 0.0959, 'widthhi': 0.1332, 'ub': 0.5661, 'lb': 0.4513})\n",
      "(26367, {'cov': 0.951, 'covstd': 0.0068, 'ubcov': 0.974, 'lbcov': 0.977, 'width': 0.0999, 'widthstd': 0.0009, 'widthlo': 0.0217, 'widthhi': 0.1199, 'ub': 0.544, 'lb': 0.4441})\n",
      "(48330, {'cov': 0.851, 'covstd': 0.0113, 'ubcov': 0.921, 'lbcov': 0.93, 'width': 0.0735, 'widthstd': 0.0012, 'widthlo': 0.0067, 'widthhi': 0.111, 'ub': 0.5399, 'lb': 0.4664})\n",
      "(88587, {'cov': 0.806, 'covstd': 0.0125, 'ubcov': 0.898, 'lbcov': 0.908, 'width': 0.049, 'widthstd': 0.0009, 'widthlo': 0.0039, 'widthhi': 0.1001, 'ub': 0.5206, 'lb': 0.4716})\n",
      "(162378, {'cov': 0.877, 'covstd': 0.0104, 'ubcov': 0.938, 'lbcov': 0.939, 'width': 0.037, 'widthstd': 0.0005, 'widthlo': 0.0043, 'widthhi': 0.058, 'ub': 0.5138, 'lb': 0.4768})\n",
      "(297636, {'cov': 0.908, 'covstd': 0.0091, 'ubcov': 0.955, 'lbcov': 0.953, 'width': 0.0276, 'widthstd': 0.0003, 'widthlo': 0.0064, 'widthhi': 0.0402, 'ub': 0.5061, 'lb': 0.4785})\n",
      "(545560, {'cov': 0.912, 'covstd': 0.009, 'ubcov': 0.966, 'lbcov': 0.946, 'width': 0.0207, 'widthstd': 0.0002, 'widthlo': 0.0073, 'widthhi': 0.0291, 'ub': 0.5112, 'lb': 0.4905})\n",
      "(1000000, {'cov': 0.927, 'covstd': 0.0082, 'ubcov': 0.972, 'lbcov': 0.955, 'width': 0.0152, 'widthstd': 0.0002, 'widthlo': 0.0054, 'widthhi': 0.0207, 'ub': 0.5144, 'lb': 0.4993})\n",
      "****** MLE+CV Bootstrap ******\n",
      "(10, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.999, 'lbcov': 0.999, 'width': 0.8568, 'widthstd': 0.0047, 'widthlo': 0.6, 'widthhi': 1.0, 'ub': 0.9297, 'lb': 0.073})\n",
      "(19, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.999, 'lbcov': 0.999, 'width': 0.7277, 'widthstd': 0.0047, 'widthlo': 0.4734, 'widthhi': 0.9474, 'ub': 0.8683, 'lb': 0.1406})\n",
      "(34, {'cov': 0.999, 'covstd': 0.001, 'ubcov': 0.999, 'lbcov': 1.0, 'width': 0.6013, 'widthstd': 0.0038, 'widthlo': 0.3823, 'widthhi': 0.7893, 'ub': 0.8093, 'lb': 0.2081})\n",
      "(62, {'cov': 1.0, 'covstd': 0.0, 'ubcov': 1.0, 'lbcov': 1.0, 'width': 0.4904, 'widthstd': 0.0031, 'widthlo': 0.3222, 'widthhi': 0.6448, 'ub': 0.7403, 'lb': 0.2498})\n",
      "(113, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.999, 'lbcov': 0.999, 'width': 0.3905, 'widthstd': 0.0025, 'widthlo': 0.2562, 'widthhi': 0.5216, 'ub': 0.6805, 'lb': 0.29})\n",
      "(207, {'cov': 0.999, 'covstd': 0.001, 'ubcov': 1.0, 'lbcov': 0.999, 'width': 0.3138, 'widthstd': 0.0019, 'widthlo': 0.2167, 'widthhi': 0.4197, 'ub': 0.6485, 'lb': 0.3347})\n",
      "(380, {'cov': 0.999, 'covstd': 0.001, 'ubcov': 1.0, 'lbcov': 0.999, 'width': 0.2565, 'widthstd': 0.0015, 'widthlo': 0.1779, 'widthhi': 0.3365, 'ub': 0.6353, 'lb': 0.3789})\n",
      "(696, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.999, 'lbcov': 0.999, 'width': 0.2103, 'widthstd': 0.0011, 'widthlo': 0.1522, 'widthhi': 0.2695, 'ub': 0.6156, 'lb': 0.4053})\n",
      "(1275, {'cov': 0.997, 'covstd': 0.0017, 'ubcov': 0.999, 'lbcov': 0.998, 'width': 0.1796, 'widthstd': 0.0009, 'widthlo': 0.1348, 'widthhi': 0.2212, 'ub': 0.6024, 'lb': 0.4228})\n",
      "(2336, {'cov': 0.997, 'covstd': 0.0017, 'ubcov': 0.997, 'lbcov': 1.0, 'width': 0.1571, 'widthstd': 0.0007, 'widthlo': 0.1207, 'widthhi': 0.1931, 'ub': 0.5808, 'lb': 0.4238})\n",
      "(4282, {'cov': 0.997, 'covstd': 0.0017, 'ubcov': 0.999, 'lbcov': 0.998, 'width': 0.1395, 'widthstd': 0.0005, 'widthlo': 0.1112, 'widthhi': 0.1673, 'ub': 0.5539, 'lb': 0.4145})\n",
      "(7848, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.998, 'lbcov': 1.0, 'width': 0.1278, 'widthstd': 0.0004, 'widthlo': 0.1061, 'widthhi': 0.1487, 'ub': 0.5591, 'lb': 0.4313})\n",
      "(14385, {'cov': 0.996, 'covstd': 0.002, 'ubcov': 0.998, 'lbcov': 0.998, 'width': 0.1169, 'widthstd': 0.0003, 'widthlo': 0.1007, 'widthhi': 0.1332, 'ub': 0.567, 'lb': 0.4501})\n",
      "(26367, {'cov': 0.998, 'covstd': 0.0014, 'ubcov': 0.999, 'lbcov': 0.999, 'width': 0.1071, 'widthstd': 0.0003, 'widthlo': 0.0905, 'widthhi': 0.1199, 'ub': 0.5474, 'lb': 0.4402})\n",
      "(48330, {'cov': 1.0, 'covstd': 0.0, 'ubcov': 1.0, 'lbcov': 1.0, 'width': 0.0948, 'widthstd': 0.0004, 'widthlo': 0.0726, 'widthhi': 0.111, 'ub': 0.5506, 'lb': 0.4558})\n",
      "(88587, {'cov': 0.995, 'covstd': 0.0022, 'ubcov': 0.997, 'lbcov': 0.998, 'width': 0.0768, 'widthstd': 0.0004, 'widthlo': 0.0528, 'widthhi': 0.1001, 'ub': 0.5351, 'lb': 0.4583})\n",
      "(162378, {'cov': 0.997, 'covstd': 0.0017, 'ubcov': 0.999, 'lbcov': 0.998, 'width': 0.055, 'widthstd': 0.0004, 'widthlo': 0.0345, 'widthhi': 0.0758, 'ub': 0.5229, 'lb': 0.4679})\n",
      "(297636, {'cov': 0.972, 'covstd': 0.0052, 'ubcov': 0.985, 'lbcov': 0.987, 'width': 0.0351, 'widthstd': 0.0003, 'widthlo': 0.0208, 'widthhi': 0.0513, 'ub': 0.5098, 'lb': 0.4748})\n",
      "(545560, {'cov': 0.905, 'covstd': 0.0093, 'ubcov': 0.963, 'lbcov': 0.942, 'width': 0.022, 'widthstd': 0.0002, 'widthlo': 0.0066, 'widthhi': 0.0322, 'ub': 0.5118, 'lb': 0.4898})\n",
      "(1000000, {'cov': 0.929, 'covstd': 0.0081, 'ubcov': 0.974, 'lbcov': 0.955, 'width': 0.0154, 'widthstd': 0.0002, 'widthlo': 0.0052, 'widthhi': 0.0212, 'ub': 0.5145, 'lb': 0.4991})\n"
     ]
    }
   ],
   "source": [
    "from importlib import reload\n",
    "import MLE.MLE\n",
    "\n",
    "reload(MLE.MLE)\n",
    "\n",
    "class CITest:\n",
    "    class ClopperPearson:\n",
    "        @staticmethod\n",
    "        def interval(data, wmax, alpha, **kwargs):    \n",
    "            from scipy.special import betaincinv\n",
    "\n",
    "            k = sum((c*w/wmax)*r for c, w, r in data)\n",
    "            n = sum(c for c, w, _ in data)\n",
    "\n",
    "            lb = betaincinv(k, n - k + 1, alpha/2) if k > 0 and n - k + 1 > 0 else 0\n",
    "            ub = betaincinv(k + 1, n - k, 1-alpha/2) if n > k else 1\n",
    "            \n",
    "            lb = min(1, max(0, wmax * lb))\n",
    "            ub = min(1, max(0, wmax * ub))\n",
    "\n",
    "            return min(lb, ub), max(lb, ub)\n",
    "        \n",
    "    class AsymptoticGaussian:\n",
    "        def interval(data, wmax, alpha, **kwargs):\n",
    "            from math import sqrt\n",
    "            from scipy.stats import norm  \n",
    "            \n",
    "            scale = norm.ppf(1-alpha/2)\n",
    " \n",
    "            n = sum(c for c, _, _ in data)\n",
    "            tot = sum(c*w*r for c, w, r in data)\n",
    "            muhat = tot / n\n",
    "            totsq = sum(c*(w*r - muhat)**2 for c, w, r in data)\n",
    "            sigmahat = sqrt(totsq) / n\n",
    "            \n",
    "            lb = min(1, max(0, muhat - scale*sigmahat))\n",
    "            ub = min(1, max(0, muhat + scale*sigmahat))\n",
    "            \n",
    "            return min(lb, ub), max(lb, ub)\n",
    "\n",
    "def produceresults(env, method, minexp=1, maxexp=6, numpts=15, ndataperpt=100):\n",
    "    from math import ceil\n",
    "    import numpy\n",
    "    \n",
    "    rangefn = env.range\n",
    "\n",
    "    for ndata in map(ceil, numpy.logspace(minexp, maxexp, numpts)):\n",
    "        bounds=[]\n",
    "        for i in range(1, ndataperpt+1):\n",
    "            (truevalue, data) = env.samplewithcvs(ndata)\n",
    "\n",
    "            ci = None\n",
    "            try:\n",
    "                ci = method(data=data, rangefn=rangefn, alpha=0.05)\n",
    "                assert numpy.isfinite(ci[0]) and numpy.isfinite(ci[1])\n",
    "                assert numpy.isfinite(ci[0]) and numpy.isfinite(ci[1])\n",
    "                assert ci[1] >= ci[0]\n",
    "                assert ci[0] >= -1e-4\n",
    "                assert ci[1] <= 1 + 1e-4\n",
    "#                 assert ci[1] >= truevalue\n",
    "#                 assert ci[0] <= truevalue\n",
    "            except:\n",
    "                print('truevalue was {}'.format(truevalue))\n",
    "                print('data was {}'.format(data))\n",
    "                print('ci was {}'.format(ci))\n",
    "                raise\n",
    "                                                \n",
    "            bounds.append(\n",
    "                ( truevalue,\n",
    "                  data,\n",
    "                  [ \n",
    "                      ci[0], 'n/a', ci[1]\n",
    "                  ]\n",
    "                )  \n",
    "            )\n",
    "            \n",
    "        uppers = [ d[2][2] for d in bounds ]\n",
    "        lowers = [ d[2][0] for d in bounds ]\n",
    "        upperbounded = [ 1 if d[0] <= d[2][2] else 0 for d in bounds ]\n",
    "        lowerbounded = [ 1 if d[2][0] <= d[0] else 0 for d in bounds ]\n",
    "        covered = [ 1 if u*l > 0 else 0 for (u, l) in zip(upperbounded, lowerbounded) ]\n",
    "        width = [ d[2][2] - d[2][0] for d in bounds ]    \n",
    "        yield (ndata,\n",
    "                { \n",
    "                    'cov': numpy.mean(covered),\n",
    "                    'covstd': numpy.std(covered, ddof=1) / numpy.sqrt(len(covered) - 1),\n",
    "                    'ubcov': numpy.mean(upperbounded),\n",
    "                    'lbcov': numpy.mean(lowerbounded),\n",
    "                    'width': numpy.mean(width),\n",
    "                    'widthstd': numpy.std(width, ddof=1) / numpy.sqrt(len(width) - 1),\n",
    "                    'widthlo': numpy.quantile(width, q=[0.05])[0],\n",
    "                    'widthhi': numpy.quantile(width, q=[0.95])[0],\n",
    "                    'ub': numpy.mean(uppers),\n",
    "                    'lb': numpy.mean(lowers),\n",
    "                },\n",
    "              )\n",
    "\n",
    "def bootstrapci(data, rangefn, alpha, **kwargs):\n",
    "    import numpy \n",
    "    from math import ceil\n",
    "    \n",
    "    what = kwargs.pop('what')\n",
    "    assert what == 'mle' or what == 'mlecv'\n",
    "    \n",
    "    p = numpy.array([ c for c, _, _, _ in data ], dtype='float64')\n",
    "    n = numpy.sum(p)\n",
    "    p /= n\n",
    "    numsamples = int(ceil(32.0 / alpha))\n",
    "    state = numpy.random.RandomState(90210)\n",
    "    resamples = state.multinomial(n=n, pvals=p, size=numsamples)\n",
    "    uppers = []\n",
    "    lowers = []\n",
    "    for counts in resamples:\n",
    "        if what == 'mle':\n",
    "            vhat, qstar = MLE.MLE.estimate(lambda: ((c, w, r) for c, (_, w, r, cvs) in zip(counts, data)),\n",
    "                                           wmin=rangefn('wmin'),\n",
    "                                           wmax=rangefn('wmax'),\n",
    "                                           raiseonerr=True)  \n",
    "        else:\n",
    "            vhat, qstar = MLE.MLE.estimatewithcv(lambda: ((c, w, r, cvs) for c, (_, w, r, cvs) in zip(counts, data)),\n",
    "                                                 rangefn=rangefn, \n",
    "                                                 raiseonerr=True)     \n",
    "        uppers.append(qstar['vmax'])\n",
    "        lowers.append(qstar['vmin'])\n",
    "        \n",
    "    ub = numpy.quantile(uppers, 1.0 - alpha / 2.0)\n",
    "    lb = numpy.quantile(lowers, alpha / 2.0)\n",
    "    \n",
    "    return lb, ub\n",
    "        \n",
    "def getenv():\n",
    "    import environments.DoubleDouble \n",
    "    wsupport = [ 0, 2, 1000 ]\n",
    "    env = environments.DoubleDouble.DoubleDouble(numactions=5, seed=45, wsupport=wsupport, expwsq=100)\n",
    "    return env\n",
    "    \n",
    "allres = []\n",
    "for (name, method) in [ \n",
    "                        ('MLE Bootstrap', lambda **kwargs: bootstrapci(what='mle', **kwargs)),\n",
    "                        ('MLE+CV Bootstrap', lambda **kwargs: bootstrapci(what='mlecv', **kwargs)),                  \n",
    "#                         ('MLE CI CV', lambda data, **kwargs: MLE.MLE.asymptoticconfidenceintervalwithcv(datagen=lambda: data, **kwargs)[0]),\n",
    "#                         ('MLE CI', lambda data, rangefn, **kwargs: MLE.MLE.asymptoticconfidenceinterval(datagen=lambda: [(c,w,r) for c, w, r, _ in data], wmin=rangefn('wmin'), wmax=rangefn('wmax'), **kwargs)[0]),\n",
    "#                         ('Asymptotic Gaussian', lambda data, rangefn, alpha, **kwargs: CITest.AsymptoticGaussian.interval([(c,w,r) for c, w, r, _ in data], rangefn('wmax'), alpha)),\n",
    "#                         ('Binomial CI', lambda data, rangefn, alpha, **kwargs: CITest.ClopperPearson.interval([(c,w,r) for c, w, r, _ in data], rangefn('wmax'), alpha)),\n",
    "                      ]:\n",
    "    print('****** {} ******'.format(name))\n",
    "    res = []\n",
    "    for zzz in produceresults(getenv(), method, ndataperpt=1000, numpts=20):\n",
    "        res.append(zzz)\n",
    "        print('{}'.format(( zzz[0], { y: round(z, 4) for y, z in zzz[1].items() })), flush=True)\n",
    "    wmax = getenv().range('wmax')\n",
    "    allres.append((name, [(x[0] / wmax, x[1]) for x in res]))\n",
    "    del wmax\n",
    "    import pickle\n",
    "    pickle.dump( allres, open( \"epsilongreedy_ciwithcv_allres_boot.p\", \"wb\" ) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "code_folding": [
     2
    ]
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "class FlassPlot:\n",
    "    from matplotlib import scale as mscale\n",
    "    \n",
    "    class ExpScale(mscale.ScaleBase):\n",
    "        from matplotlib import transforms as mtransforms\n",
    "\n",
    "        \"\"\"\n",
    "        Scales data in range -Infinity, 1 using log(1 - x)\n",
    "        \"\"\"\n",
    "\n",
    "        # The scale class must have a member ``name`` that defines the\n",
    "        # string used to select the scale.  For example,\n",
    "        # ``gca().set_yscale(\"mercator\")`` would be used to select this\n",
    "        # scale.\n",
    "        name = 'exp'\n",
    "\n",
    "        def __init__(self, axis, **kwargs):\n",
    "            \"\"\"\n",
    "            Any keyword arguments passed to ``set_xscale`` and\n",
    "            ``set_yscale`` will be passed along to the scale's\n",
    "            constructor.\n",
    "            \"\"\"\n",
    "            from matplotlib import scale as mscale\n",
    "            myscale = kwargs.pop(\"expscale\", 1.0)\n",
    "            mscale.ScaleBase.__init__(self, axis, **kwargs)\n",
    "            self.myscale = myscale\n",
    "\n",
    "        def get_transform(self):\n",
    "            \"\"\"\n",
    "            Override this method to return a new instance that does the\n",
    "            actual transformation of the data.\n",
    "            \"\"\"\n",
    "            return FlassPlot.ExpScale.ExpTransform(self.myscale)\n",
    "\n",
    "        def set_default_locators_and_formatters(self, axis):\n",
    "            \"\"\"\n",
    "            Override to set up the locators and formatters to use with the\n",
    "            scale.  This is only required if the scale requires custom\n",
    "            locators and formatters.  Writing custom locators and\n",
    "            formatters is rather outside the scope of this example, but\n",
    "            there are many helpful examples in ``ticker.py``.\n",
    "\n",
    "            In our case, the Mercator example uses a fixed locator from\n",
    "            -90 to 90 degrees and a custom formatter class to put convert\n",
    "            the radians to degrees and put a degree symbol after the\n",
    "            value::\n",
    "            \"\"\"\n",
    "\n",
    "            return axis\n",
    "\n",
    "        def limit_range_for_scale(self, vmin, vmax, minpos):\n",
    "            \"\"\"\n",
    "            Override to limit the bounds of the axis to the domain of the\n",
    "            transform.\n",
    "            \"\"\"\n",
    "            return vmin, vmax\n",
    "\n",
    "        class ExpTransform(mtransforms.Transform):\n",
    "            # There are two value members that must be defined.\n",
    "            # ``input_dims`` and ``output_dims`` specify number of input\n",
    "            # dimensions and output dimensions to the transformation.\n",
    "            # These are used by the transformation framework to do some\n",
    "            # error checking and prevent incompatible transformations from\n",
    "            # being connected together.  When defining transforms for a\n",
    "            # scale, which are, by definition, separable and have only one\n",
    "            # dimension, these members should always be set to 1.\n",
    "            input_dims = 1\n",
    "            output_dims = 1\n",
    "            is_separable = True\n",
    "\n",
    "            def __init__(self, scale):\n",
    "                from matplotlib import transforms as mtransforms\n",
    "                \n",
    "                mtransforms.Transform.__init__(self)\n",
    "                self.scale = scale\n",
    "\n",
    "            def transform_non_affine(self, a):\n",
    "                \"\"\"\n",
    "                This transform takes an Nx1 ``numpy`` array and returns a\n",
    "                transformed copy.  Since the range of the Mercator scale\n",
    "                is limited by the user-specified threshold, the input\n",
    "                array must be masked to contain only valid values.\n",
    "                ``matplotlib`` will handle masked arrays and remove the\n",
    "                out-of-range data from the plot.  Importantly, the\n",
    "                ``transform`` method *must* return an array that is the\n",
    "                same shape as the input array, since these values need to\n",
    "                remain synchronized with values in the other dimension.\n",
    "                \"\"\"\n",
    "                import numpy as np\n",
    "                \n",
    "                return np.exp(self.scale * a)\n",
    "\n",
    "            def inverted(self):\n",
    "                \"\"\"\n",
    "                Override this method so matplotlib knows how to get the\n",
    "                inverse transform for this transform.\n",
    "                \"\"\"\n",
    "                return FlassPlot.ExpScale.InvertedExpTransform(self.scale)\n",
    "\n",
    "        class InvertedExpTransform(mtransforms.Transform):\n",
    "            input_dims = 1\n",
    "            output_dims = 1\n",
    "            is_separable = True\n",
    "\n",
    "            def __init__(self, scale):\n",
    "                from matplotlib import transforms as mtransforms\n",
    "                \n",
    "                mtransforms.Transform.__init__(self)\n",
    "                self.scale = scale\n",
    "\n",
    "            def transform_non_affine(self, a):\n",
    "                import numpy as np\n",
    "                \n",
    "                return np.log(a / self.scale)\n",
    "\n",
    "            def inverted(self):\n",
    "                return FlassPlot.ExpScale.ExpTransform(self.scale)\n",
    "\n",
    "    @staticmethod\n",
    "    def yo():\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.figure(figsize=(4,2))\n",
    "        \n",
    "    @staticmethod\n",
    "    def forpaper():\n",
    "        import matplotlib.pyplot as plt\n",
    "        from matplotlib import scale as mscale\n",
    "\n",
    "        mscale.register_scale(FlassPlot.ExpScale)\n",
    "\n",
    "        SMALL_SIZE = 16\n",
    "        MEDIUM_SIZE = 22\n",
    "        BIGGER_SIZE = 24\n",
    "\n",
    "        plt.rc('font', size=SMALL_SIZE)          # controls default text sizes\n",
    "        plt.rc('axes', titlesize=SMALL_SIZE)     # fontsize of the axes title\n",
    "        plt.rc('axes', labelsize=MEDIUM_SIZE)    # fontsize of the x and y labels\n",
    "        plt.rc('xtick', labelsize=SMALL_SIZE)    # fontsize of the tick labels\n",
    "        plt.rc('ytick', labelsize=SMALL_SIZE)    # fontsize of the tick labels\n",
    "        plt.rc('legend', fontsize=SMALL_SIZE)    # legend fontsize\n",
    "        plt.rc('figure', titlesize=BIGGER_SIZE)  # fontsize of the figure title\n",
    "        \n",
    "    @staticmethod\n",
    "    def exppic(x, y, label, expscale, **kwargs):\n",
    "        import matplotlib.pyplot as plt\n",
    "        import matplotlib.ticker as ticker\n",
    "        plt.ylim(0.0, 1.01)\n",
    "        plt.plot(x, y, label=label, **kwargs)\n",
    "        plt.gca().set_xscale('log')\n",
    "        plt.gca().set_yscale('exp', expscale=expscale)\n",
    "        plt.ylim(0.0, 1.01)\n",
    "        plt.gca().yaxis.set_major_locator(ticker.FixedLocator([0.38,0.68,0.86,0.95,1]))\n",
    "\n",
    "        plt.legend()\n",
    "        \n",
    "    @staticmethod\n",
    "    def pic(x, y, label): \n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.semilogx(x, y, label=label)\n",
    "\n",
    "        #plt.loglog(x, y, label=label)\n",
    "        plt.legend()\n",
    "        \n",
    "    @staticmethod\n",
    "    def axeslabel(xlabel, ylabel):  \n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.xlabel(xlabel)\n",
    "        plt.ylabel(ylabel)\n",
    "        \n",
    "    @staticmethod\n",
    "    def title(title):\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.title(title)\n",
    "        \n",
    "    @staticmethod\n",
    "    def legend(**kwargs):\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.legend(**kwargs)\n",
    "        \n",
    "    @staticmethod\n",
    "    def savefig(filename):\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.savefig(filename, bbox_inches='tight')\n",
    "        \n",
    "    @staticmethod\n",
    "    def plt():\n",
    "        import matplotlib.pyplot as plt\n",
    "        return plt\n",
    "\n",
    "import pickle\n",
    "del allres\n",
    "allres = pickle.load(open( \"epsilongreedy_ciwithcv_allres_big.p\", \"rb\" ) )\n",
    "megallres = pickle.load(open( \"epsilongreedy_ciwithcv_allres_boot.p\", \"rb\" ) )\n",
    "allres.extend(megallres)\n",
    "\n",
    "renameit = { \"Binomial CI\": \"Binomial\", \"Asymptotic Gaussian\": \"Gaussian\", \"MLE CI\": \"MLE\", \"MLE CI CV\": \"MLE+cv\" }\n",
    "skip = { \"Binomial CI\": \"Binomial\", \"Asymptotic Gaussian\": \"Gaussian\", }\n",
    "FlassPlot.forpaper()\n",
    "for name, res in reversed(allres):\n",
    "    if name in skip:\n",
    "        continue\n",
    "    x = [ x[0] for x in res ]\n",
    "    y = [ x[1]['width'] for x in res ]\n",
    "    ylo = [ x[1]['widthlo'] for x in res ]\n",
    "    yhi = [ x[1]['widthhi'] for x in res ]\n",
    "    #ylo = [ x[1]['width'] - 1.96 * x[1]['widthstd'] for x in res ]\n",
    "    #yhi = [ x[1]['width'] + 1.96 * x[1]['widthstd'] for x in res ]\n",
    "    FlassPlot.plt().loglog(x, y, label=renameit.get(name, name))\n",
    "    FlassPlot.plt().fill_between(x, ylo, yhi, alpha=0.15)\n",
    "\n",
    "FlassPlot.legend()#loc=\"center\", bbox_to_anchor=[0.8,0.825])\n",
    "FlassPlot.axeslabel('n / $w_{max}$', 'CI width')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import pickle\n",
    "del allres\n",
    "allres = pickle.load(open( \"epsilongreedy_ciwithcv_allres_big.p\", \"rb\" ) )\n",
    "megallres = pickle.load(open( \"epsilongreedy_ciwithcv_allres_boot.p\", \"rb\" ) )\n",
    "allres.extend(megallres)\n",
    "\n",
    "renameit = { \"Binomial CI\": \"Binomial\", \"Asymptotic Gaussian\": \"Gaussian\", \"MLE CI\": \"MLE\", \"MLE CI CV\": \"MLE+cv\" }\n",
    "skip = { \"Binomial CI\": \"Binomial\", \"Asymptotic Gaussian\": \"Gaussian\", }\n",
    "FlassPlot.forpaper()\n",
    "for name, res in reversed(allres):\n",
    "    if name in skip:\n",
    "        continue\n",
    "    x = [ x[0] for x in res ]\n",
    "    y = [ x[1]['cov'] for x in res ]\n",
    "    ylo = [ x[1]['cov'] - 1.96 * x[1]['covstd'] for x in res ]\n",
    "    yhi = [ x[1]['cov'] + 1.96 * x[1]['covstd'] for x in res ]\n",
    "    FlassPlot.exppic(x, y, label=renameit.get(name, name), expscale=7.0)\n",
    "    FlassPlot.plt().fill_between(x, ylo, yhi, alpha=0.25)\n",
    "FlassPlot.exppic([ x[0] for x in res ], [ 0.95 for x in res ], label='_nolegend_', expscale=7.0, ls='dashed')\n",
    "FlassPlot.legend()#loc=\"center\", bbox_to_anchor=[0.5,0.425])\n",
    "FlassPlot.axeslabel('n / $w_{max}$', 'empirical coverage')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "code_folding": [
     6,
     22,
     40
    ],
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "****** MLE CI CV ******\n",
      "(10, {'cov': 0.9989, 'covstd': 0.0001, 'ubcov': 0.9995, 'lbcov': 0.9994, 'width': 0.8159, 'widthstd': 0.0005, 'widthlo': 0.5842, 'widthhi': 0.9717, 'ub': 0.9078, 'lb': 0.0919})\n",
      "(19, {'cov': 0.9982, 'covstd': 0.0002, 'ubcov': 0.9991, 'lbcov': 0.9991, 'width': 0.6792, 'widthstd': 0.0006, 'widthlo': 0.4568, 'widthhi': 0.8718, 'ub': 0.8395, 'lb': 0.1603})\n",
      "(34, {'cov': 0.9976, 'covstd': 0.0002, 'ubcov': 0.9989, 'lbcov': 0.9987, 'width': 0.5621, 'widthstd': 0.0005, 'widthlo': 0.3678, 'widthhi': 0.7456, 'ub': 0.7809, 'lb': 0.2188})\n",
      "(62, {'cov': 0.9973, 'covstd': 0.0002, 'ubcov': 0.9985, 'lbcov': 0.9988, 'width': 0.4557, 'widthstd': 0.0004, 'widthlo': 0.2962, 'widthhi': 0.6138, 'ub': 0.7277, 'lb': 0.272})\n",
      "(113, {'cov': 0.9974, 'covstd': 0.0002, 'ubcov': 0.9989, 'lbcov': 0.9985, 'width': 0.3683, 'widthstd': 0.0003, 'widthlo': 0.2429, 'widthhi': 0.4941, 'ub': 0.6826, 'lb': 0.3142})\n",
      "(207, {'cov': 0.9966, 'covstd': 0.0003, 'ubcov': 0.9984, 'lbcov': 0.9982, 'width': 0.2986, 'widthstd': 0.0003, 'widthlo': 0.1994, 'widthhi': 0.3992, 'ub': 0.6492, 'lb': 0.3506})\n",
      "(380, {'cov': 0.9967, 'covstd': 0.0003, 'ubcov': 0.9983, 'lbcov': 0.9984, 'width': 0.2456, 'widthstd': 0.0002, 'widthlo': 0.1677, 'widthhi': 0.3235, 'ub': 0.6233, 'lb': 0.3777})\n",
      "(696, {'cov': 0.9964, 'covstd': 0.0003, 'ubcov': 0.9981, 'lbcov': 0.9983, 'width': 0.2051, 'widthstd': 0.0002, 'widthlo': 0.1447, 'widthhi': 0.2654, 'ub': 0.6012, 'lb': 0.3961})\n",
      "(1275, {'cov': 0.9964, 'covstd': 0.0003, 'ubcov': 0.9984, 'lbcov': 0.998, 'width': 0.1755, 'widthstd': 0.0001, 'widthlo': 0.1286, 'widthhi': 0.2224, 'ub': 0.587, 'lb': 0.4114})\n",
      "(2336, {'cov': 0.9969, 'covstd': 0.0002, 'ubcov': 0.9984, 'lbcov': 0.9985, 'width': 0.1532, 'widthstd': 0.0001, 'widthlo': 0.1163, 'widthhi': 0.1892, 'ub': 0.5759, 'lb': 0.4226})\n",
      "(4282, {'cov': 0.9968, 'covstd': 0.0003, 'ubcov': 0.9984, 'lbcov': 0.9984, 'width': 0.1358, 'widthstd': 0.0001, 'widthlo': 0.1063, 'widthhi': 0.1643, 'ub': 0.5661, 'lb': 0.4303})\n",
      "(7848, {'cov': 0.9963, 'covstd': 0.0003, 'ubcov': 0.9982, 'lbcov': 0.9982, 'width': 0.121, 'widthstd': 0.0001, 'widthlo': 0.0949, 'widthhi': 0.1447, 'ub': 0.56, 'lb': 0.439})\n",
      "(14385, {'cov': 0.9957, 'covstd': 0.0003, 'ubcov': 0.998, 'lbcov': 0.9977, 'width': 0.107, 'widthstd': 0.0001, 'widthlo': 0.0806, 'widthhi': 0.1293, 'ub': 0.5532, 'lb': 0.4462})\n",
      "(26367, {'cov': 0.991, 'covstd': 0.0004, 'ubcov': 0.9956, 'lbcov': 0.9954, 'width': 0.0914, 'widthstd': 0.0001, 'widthlo': 0.0626, 'widthhi': 0.1155, 'ub': 0.5448, 'lb': 0.4535})\n",
      "(48330, {'cov': 0.9773, 'covstd': 0.0007, 'ubcov': 0.9888, 'lbcov': 0.9885, 'width': 0.0731, 'widthstd': 0.0001, 'widthlo': 0.0448, 'widthhi': 0.1007, 'ub': 0.5338, 'lb': 0.4607})\n",
      "(88587, {'cov': 0.9549, 'covstd': 0.0009, 'ubcov': 0.978, 'lbcov': 0.9769, 'width': 0.0532, 'widthstd': 0.0001, 'widthlo': 0.0294, 'widthhi': 0.0793, 'ub': 0.5275, 'lb': 0.4743})\n",
      "(162378, {'cov': 0.9339, 'covstd': 0.0011, 'ubcov': 0.9681, 'lbcov': 0.9659, 'width': 0.037, 'widthstd': 0.0, 'widthlo': 0.0193, 'widthhi': 0.0563, 'ub': 0.5191, 'lb': 0.4821})\n",
      "(297636, {'cov': 0.9319, 'covstd': 0.0011, 'ubcov': 0.9655, 'lbcov': 0.9664, 'width': 0.0268, 'widthstd': 0.0, 'widthlo': 0.0124, 'widthhi': 0.0379, 'ub': 0.5105, 'lb': 0.4837})\n",
      "(545560, {'cov': 0.9379, 'covstd': 0.0011, 'ubcov': 0.9685, 'lbcov': 0.9694, 'width': 0.0201, 'widthstd': 0.0, 'widthlo': 0.0081, 'widthhi': 0.0277, 'ub': 0.5101, 'lb': 0.49})\n",
      "(1000000, {'cov': 0.943, 'covstd': 0.001, 'ubcov': 0.971, 'lbcov': 0.972, 'width': 0.015, 'widthstd': 0.0, 'widthlo': 0.0058, 'widthhi': 0.0202, 'ub': 0.5097, 'lb': 0.4947})\n",
      "****** MLE CI ******\n",
      "(10, {'cov': 0.9964, 'covstd': 0.0003, 'ubcov': 0.9984, 'lbcov': 0.998, 'width': 0.7703, 'widthstd': 0.0007, 'widthlo': 0.502, 'widthhi': 0.9495, 'ub': 0.8848, 'lb': 0.1145})\n",
      "(19, {'cov': 0.9945, 'covstd': 0.0003, 'ubcov': 0.9974, 'lbcov': 0.997, 'width': 0.6344, 'widthstd': 0.0007, 'widthlo': 0.3722, 'widthhi': 0.8672, 'ub': 0.817, 'lb': 0.1826})\n",
      "(34, {'cov': 0.9936, 'covstd': 0.0004, 'ubcov': 0.9971, 'lbcov': 0.9965, 'width': 0.5242, 'widthstd': 0.0006, 'widthlo': 0.2988, 'widthhi': 0.734, 'ub': 0.762, 'lb': 0.2378})\n",
      "(62, {'cov': 0.9929, 'covstd': 0.0004, 'ubcov': 0.9964, 'lbcov': 0.9965, 'width': 0.4265, 'widthstd': 0.0005, 'widthlo': 0.2492, 'widthhi': 0.6047, 'ub': 0.7131, 'lb': 0.2866})\n",
      "(113, {'cov': 0.993, 'covstd': 0.0004, 'ubcov': 0.9967, 'lbcov': 0.9963, 'width': 0.3481, 'widthstd': 0.0004, 'widthlo': 0.2089, 'widthhi': 0.4871, 'ub': 0.6724, 'lb': 0.3243})\n",
      "(207, {'cov': 0.9921, 'covstd': 0.0004, 'ubcov': 0.9961, 'lbcov': 0.9959, 'width': 0.2858, 'widthstd': 0.0003, 'widthlo': 0.1752, 'widthhi': 0.3959, 'ub': 0.6428, 'lb': 0.357})\n",
      "(380, {'cov': 0.9931, 'covstd': 0.0004, 'ubcov': 0.9967, 'lbcov': 0.9964, 'width': 0.2381, 'widthstd': 0.0002, 'widthlo': 0.1515, 'widthhi': 0.3225, 'ub': 0.6196, 'lb': 0.3815})\n",
      "(696, {'cov': 0.9938, 'covstd': 0.0004, 'ubcov': 0.9967, 'lbcov': 0.9971, 'width': 0.2007, 'widthstd': 0.0002, 'widthlo': 0.1341, 'widthhi': 0.2648, 'ub': 0.599, 'lb': 0.3983})\n",
      "(1275, {'cov': 0.9935, 'covstd': 0.0004, 'ubcov': 0.9969, 'lbcov': 0.9966, 'width': 0.1718, 'widthstd': 0.0001, 'widthlo': 0.1189, 'widthhi': 0.222, 'ub': 0.5851, 'lb': 0.4133})\n",
      "(2336, {'cov': 0.9923, 'covstd': 0.0004, 'ubcov': 0.9959, 'lbcov': 0.9964, 'width': 0.1486, 'widthstd': 0.0001, 'widthlo': 0.1034, 'widthhi': 0.1889, 'ub': 0.5735, 'lb': 0.4249})\n",
      "(4282, {'cov': 0.988, 'covstd': 0.0005, 'ubcov': 0.9939, 'lbcov': 0.9942, 'width': 0.1291, 'widthstd': 0.0001, 'widthlo': 0.0886, 'widthhi': 0.1639, 'ub': 0.5628, 'lb': 0.4337})\n",
      "(7848, {'cov': 0.9829, 'covstd': 0.0006, 'ubcov': 0.9914, 'lbcov': 0.9914, 'width': 0.1113, 'widthstd': 0.0001, 'widthlo': 0.0737, 'widthhi': 0.1441, 'ub': 0.5551, 'lb': 0.4438})\n",
      "(14385, {'cov': 0.9722, 'covstd': 0.0007, 'ubcov': 0.986, 'lbcov': 0.9862, 'width': 0.0941, 'widthstd': 0.0001, 'widthlo': 0.0584, 'widthhi': 0.1284, 'ub': 0.5468, 'lb': 0.4527})\n",
      "(26367, {'cov': 0.9603, 'covstd': 0.0009, 'ubcov': 0.98, 'lbcov': 0.9803, 'width': 0.0765, 'widthstd': 0.0001, 'widthlo': 0.0431, 'widthhi': 0.1125, 'ub': 0.5374, 'lb': 0.4609})\n",
      "(48330, {'cov': 0.9489, 'covstd': 0.001, 'ubcov': 0.9738, 'lbcov': 0.9751, 'width': 0.0604, 'widthstd': 0.0001, 'widthlo': 0.0308, 'widthhi': 0.0903, 'ub': 0.5274, 'lb': 0.467})\n",
      "(88587, {'cov': 0.9463, 'covstd': 0.001, 'ubcov': 0.9737, 'lbcov': 0.9726, 'width': 0.0469, 'widthstd': 0.0001, 'widthlo': 0.0208, 'widthhi': 0.0697, 'ub': 0.5244, 'lb': 0.4775})\n",
      "(162378, {'cov': 0.9454, 'covstd': 0.001, 'ubcov': 0.9732, 'lbcov': 0.9722, 'width': 0.036, 'widthstd': 0.0001, 'widthlo': 0.0137, 'widthhi': 0.0518, 'ub': 0.5186, 'lb': 0.4826})\n",
      "(297636, {'cov': 0.948, 'covstd': 0.001, 'ubcov': 0.9737, 'lbcov': 0.9743, 'width': 0.0271, 'widthstd': 0.0, 'widthlo': 0.0095, 'widthhi': 0.0379, 'ub': 0.5106, 'lb': 0.4835})\n",
      "(545560, {'cov': 0.9474, 'covstd': 0.001, 'ubcov': 0.9733, 'lbcov': 0.9741, 'width': 0.0202, 'widthstd': 0.0, 'widthlo': 0.0074, 'widthhi': 0.0275, 'ub': 0.5101, 'lb': 0.49})\n",
      "(1000000, {'cov': 0.9468, 'covstd': 0.001, 'ubcov': 0.9733, 'lbcov': 0.9735, 'width': 0.015, 'widthstd': 0.0, 'widthlo': 0.0059, 'widthhi': 0.02, 'ub': 0.5097, 'lb': 0.4947})\n",
      "****** Asymptotic Gaussian ******\n",
      "(10, {'cov': 0.7199, 'covstd': 0.002, 'ubcov': 0.7328, 'lbcov': 0.9871, 'width': 0.5778, 'widthstd': 0.0016, 'widthlo': 0.0, 'widthhi': 0.9681, 'ub': 0.6757, 'lb': 0.098})\n",
      "(19, {'cov': 0.7849, 'covstd': 0.0018, 'ubcov': 0.7936, 'lbcov': 0.9913, 'width': 0.5113, 'widthstd': 0.0012, 'widthlo': 0.0, 'widthhi': 0.792, 'ub': 0.6685, 'lb': 0.1571})\n",
      "(34, {'cov': 0.8097, 'covstd': 0.0018, 'ubcov': 0.817, 'lbcov': 0.9927, 'width': 0.4227, 'widthstd': 0.0009, 'widthlo': 0.0, 'widthhi': 0.629, 'ub': 0.6396, 'lb': 0.2169})\n",
      "(62, {'cov': 0.8123, 'covstd': 0.0017, 'ubcov': 0.8183, 'lbcov': 0.994, 'width': 0.3359, 'widthstd': 0.0006, 'widthlo': 0.0, 'widthhi': 0.4764, 'ub': 0.6074, 'lb': 0.2715})\n",
      "(113, {'cov': 0.7957, 'covstd': 0.0018, 'ubcov': 0.8001, 'lbcov': 0.9956, 'width': 0.2622, 'widthstd': 0.0005, 'widthlo': 0.0522, 'widthhi': 0.3596, 'ub': 0.575, 'lb': 0.3128})\n",
      "(207, {'cov': 0.7439, 'covstd': 0.002, 'ubcov': 0.7481, 'lbcov': 0.9958, 'width': 0.2043, 'widthstd': 0.0005, 'widthlo': 0.046, 'widthhi': 0.2681, 'ub': 0.5513, 'lb': 0.3469})\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(380, {'cov': 0.6592, 'covstd': 0.0021, 'ubcov': 0.6619, 'lbcov': 0.9973, 'width': 0.1631, 'widthstd': 0.0006, 'widthlo': 0.0336, 'widthhi': 0.1993, 'ub': 0.533, 'lb': 0.3699})\n",
      "(696, {'cov': 0.556, 'covstd': 0.0022, 'ubcov': 0.5579, 'lbcov': 0.9981, 'width': 0.1384, 'widthstd': 0.0007, 'widthlo': 0.0275, 'widthhi': 0.1479, 'ub': 0.5194, 'lb': 0.381})\n",
      "(1275, {'cov': 0.4567, 'covstd': 0.0022, 'ubcov': 0.4586, 'lbcov': 0.9981, 'width': 0.1356, 'widthstd': 0.001, 'widthlo': 0.0203, 'widthhi': 0.938, 'ub': 0.5197, 'lb': 0.3841})\n",
      "(2336, {'cov': 0.3947, 'covstd': 0.0022, 'ubcov': 0.3961, 'lbcov': 0.9987, 'width': 0.1406, 'widthstd': 0.0011, 'widthlo': 0.016, 'widthhi': 0.9299, 'ub': 0.5307, 'lb': 0.3901})\n",
      "(4282, {'cov': 0.3938, 'covstd': 0.0022, 'ubcov': 0.3949, 'lbcov': 0.9989, 'width': 0.1585, 'widthstd': 0.0011, 'widthlo': 0.0122, 'widthhi': 0.8239, 'ub': 0.5547, 'lb': 0.3962})\n",
      "(7848, {'cov': 0.4525, 'covstd': 0.0022, 'ubcov': 0.4537, 'lbcov': 0.9989, 'width': 0.1601, 'widthstd': 0.0009, 'widthlo': 0.0099, 'widthhi': 0.5008, 'ub': 0.567, 'lb': 0.4069})\n",
      "(14385, {'cov': 0.5708, 'covstd': 0.0022, 'ubcov': 0.5715, 'lbcov': 0.9993, 'width': 0.1484, 'widthstd': 0.0007, 'widthlo': 0.0083, 'widthhi': 0.3862, 'ub': 0.5671, 'lb': 0.4187})\n",
      "(26367, {'cov': 0.7141, 'covstd': 0.002, 'ubcov': 0.7156, 'lbcov': 0.9985, 'width': 0.1284, 'widthstd': 0.0004, 'widthlo': 0.0074, 'widthhi': 0.2974, 'ub': 0.5598, 'lb': 0.4315})\n",
      "(48330, {'cov': 0.799, 'covstd': 0.0018, 'ubcov': 0.8016, 'lbcov': 0.9974, 'width': 0.1039, 'widthstd': 0.0003, 'widthlo': 0.0077, 'widthhi': 0.1991, 'ub': 0.547, 'lb': 0.4431})\n",
      "(88587, {'cov': 0.8587, 'covstd': 0.0016, 'ubcov': 0.8632, 'lbcov': 0.9955, 'width': 0.081, 'widthstd': 0.0002, 'widthlo': 0.0088, 'widthhi': 0.1402, 'ub': 0.5403, 'lb': 0.4592})\n",
      "(162378, {'cov': 0.897, 'covstd': 0.0014, 'ubcov': 0.9031, 'lbcov': 0.9939, 'width': 0.0616, 'widthstd': 0.0001, 'widthlo': 0.0088, 'widthhi': 0.0998, 'ub': 0.5307, 'lb': 0.4691})\n",
      "(297636, {'cov': 0.9176, 'covstd': 0.0012, 'ubcov': 0.9261, 'lbcov': 0.9915, 'width': 0.0462, 'widthstd': 0.0001, 'widthlo': 0.0134, 'widthhi': 0.0723, 'ub': 0.5198, 'lb': 0.4736})\n",
      "(545560, {'cov': 0.9307, 'covstd': 0.0011, 'ubcov': 0.942, 'lbcov': 0.9887, 'width': 0.0345, 'widthstd': 0.0001, 'widthlo': 0.0103, 'widthhi': 0.0525, 'ub': 0.517, 'lb': 0.4825})\n",
      "(1000000, {'cov': 0.9373, 'covstd': 0.0011, 'ubcov': 0.9514, 'lbcov': 0.9859, 'width': 0.0256, 'widthstd': 0.0, 'widthlo': 0.0078, 'widthhi': 0.0384, 'ub': 0.5149, 'lb': 0.4892})\n",
      "****** Binomial CI ******\n",
      "(10, {'cov': 0.9995, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9995, 'width': 0.9995, 'widthstd': 0.0001, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0005})\n",
      "(19, {'cov': 0.999, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.999, 'width': 0.999, 'widthstd': 0.0001, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.001})\n",
      "(34, {'cov': 0.9988, 'covstd': 0.0002, 'ubcov': 1.0, 'lbcov': 0.9988, 'width': 0.9986, 'widthstd': 0.0002, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0014})\n",
      "(62, {'cov': 0.9986, 'covstd': 0.0002, 'ubcov': 1.0, 'lbcov': 0.9986, 'width': 0.9986, 'widthstd': 0.0001, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0014})\n",
      "(113, {'cov': 0.9988, 'covstd': 0.0002, 'ubcov': 1.0, 'lbcov': 0.9988, 'width': 0.9984, 'widthstd': 0.0001, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0016})\n",
      "(207, {'cov': 0.9992, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9992, 'width': 0.9981, 'widthstd': 0.0001, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0019})\n",
      "(380, {'cov': 0.9994, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9994, 'width': 0.9975, 'widthstd': 0.0001, 'widthlo': 1.0, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0025})\n",
      "(696, {'cov': 0.9996, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9996, 'width': 0.9956, 'widthstd': 0.0001, 'widthlo': 0.9962, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0044})\n",
      "(1275, {'cov': 0.9996, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9996, 'width': 0.9853, 'widthstd': 0.0002, 'widthlo': 0.9563, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0147})\n",
      "(2336, {'cov': 0.9993, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9993, 'width': 0.9557, 'widthstd': 0.0003, 'widthlo': 0.858, 'widthhi': 1.0, 'ub': 1.0, 'lb': 0.0443})\n",
      "(4282, {'cov': 0.9996, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9996, 'width': 0.8949, 'widthstd': 0.0004, 'widthlo': 0.7227, 'widthhi': 0.9998, 'ub': 0.9912, 'lb': 0.0964})\n",
      "(7848, {'cov': 0.9993, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9993, 'width': 0.7577, 'widthstd': 0.0006, 'widthlo': 0.5036, 'widthhi': 0.9368, 'ub': 0.9178, 'lb': 0.16})\n",
      "(14385, {'cov': 0.9994, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9994, 'width': 0.603, 'widthstd': 0.0006, 'widthlo': 0.3184, 'widthhi': 0.7892, 'ub': 0.8279, 'lb': 0.2249})\n",
      "(26367, {'cov': 0.9994, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9994, 'width': 0.4644, 'widthstd': 0.0006, 'widthlo': 0.2168, 'widthhi': 0.6294, 'ub': 0.7475, 'lb': 0.2831})\n",
      "(48330, {'cov': 0.9994, 'covstd': 0.0001, 'ubcov': 1.0, 'lbcov': 0.9994, 'width': 0.3524, 'widthstd': 0.0005, 'widthlo': 0.1482, 'widthhi': 0.4886, 'ub': 0.6829, 'lb': 0.3304})\n",
      "(88587, {'cov': 0.9995, 'covstd': 0.0001, 'ubcov': 0.9999, 'lbcov': 0.9996, 'width': 0.266, 'widthstd': 0.0004, 'widthlo': 0.1058, 'widthhi': 0.3732, 'ub': 0.6391, 'lb': 0.3731})\n",
      "(162378, {'cov': 0.999, 'covstd': 0.0001, 'ubcov': 0.9998, 'lbcov': 0.9992, 'width': 0.1988, 'widthstd': 0.0003, 'widthlo': 0.0759, 'widthhi': 0.2824, 'ub': 0.6028, 'lb': 0.404})\n",
      "(297636, {'cov': 0.9991, 'covstd': 0.0001, 'ubcov': 0.9996, 'lbcov': 0.9994, 'width': 0.1479, 'widthstd': 0.0002, 'widthlo': 0.0535, 'widthhi': 0.2119, 'ub': 0.5728, 'lb': 0.4249})\n",
      "(545560, {'cov': 0.9986, 'covstd': 0.0002, 'ubcov': 0.9995, 'lbcov': 0.9992, 'width': 0.11, 'widthstd': 0.0002, 'widthlo': 0.0393, 'widthhi': 0.1584, 'ub': 0.5558, 'lb': 0.4459})\n",
      "(1000000, {'cov': 0.9986, 'covstd': 0.0002, 'ubcov': 0.9995, 'lbcov': 0.9991, 'width': 0.0818, 'widthstd': 0.0001, 'widthlo': 0.0285, 'widthhi': 0.1179, 'ub': 0.5435, 'lb': 0.4617})\n"
     ]
    }
   ],
   "source": [
    "from importlib import reload\n",
    "import MLE.MLE\n",
    "\n",
    "reload(MLE.MLE)\n",
    "\n",
    "class CITest:\n",
    "    class ClopperPearson:\n",
    "        @staticmethod\n",
    "        def interval(data, wmax, alpha, **kwargs):    \n",
    "            from scipy.special import betaincinv\n",
    "\n",
    "            k = sum((c*w/wmax)*r for c, w, r in data)\n",
    "            n = sum(c for c, w, _ in data)\n",
    "\n",
    "            lb = betaincinv(k, n - k + 1, alpha/2) if k > 0 and n - k + 1 > 0 else 0\n",
    "            ub = betaincinv(k + 1, n - k, 1-alpha/2) if n > k else 1\n",
    "            \n",
    "            lb = min(1, max(0, wmax * lb))\n",
    "            ub = min(1, max(0, wmax * ub))\n",
    "\n",
    "            return min(lb, ub), max(lb, ub)\n",
    "        \n",
    "    class AsymptoticGaussian:\n",
    "        def interval(data, wmax, alpha, **kwargs):\n",
    "            from math import sqrt\n",
    "            from scipy.stats import norm  \n",
    "            \n",
    "            scale = norm.ppf(1-alpha/2)\n",
    " \n",
    "            n = sum(c for c, _, _ in data)\n",
    "            tot = sum(c*w*r for c, w, r in data)\n",
    "            muhat = tot / n\n",
    "            totsq = sum(c*(w*r - muhat)**2 for c, w, r in data)\n",
    "            sigmahat = sqrt(totsq) / n\n",
    "            \n",
    "            lb = min(1, max(0, muhat - scale*sigmahat))\n",
    "            ub = min(1, max(0, muhat + scale*sigmahat))\n",
    "            \n",
    "            return min(lb, ub), max(lb, ub)\n",
    "\n",
    "def produceresults(env, method, minexp=1, maxexp=6, numpts=15, ndataperpt=100):\n",
    "    from math import ceil\n",
    "    import numpy\n",
    "    \n",
    "    rangefn = env.range\n",
    "\n",
    "    for ndata in map(ceil, numpy.logspace(minexp, maxexp, numpts)):\n",
    "        bounds=[]\n",
    "        for i in range(1, ndataperpt+1):\n",
    "            (truevalue, data) = env.samplewithcvs(ndata)\n",
    "\n",
    "            try:\n",
    "                ci = method(data=data, rangefn=rangefn, alpha=0.05)\n",
    "                assert numpy.isfinite(ci[0]) and numpy.isfinite(ci[1])\n",
    "                assert numpy.isfinite(ci[0]) and numpy.isfinite(ci[1])\n",
    "                assert ci[1] >= ci[0]\n",
    "                assert ci[0] >= -1e-4\n",
    "                assert ci[1] <= 1 + 1e-4\n",
    "#                 assert ci[1] >= truevalue\n",
    "#                 assert ci[0] <= truevalue\n",
    "            except:\n",
    "                print('truevalue was {}'.format(truevalue))\n",
    "                print('data was {}'.format(data))\n",
    "                print('ci was {}'.format(ci))\n",
    "                raise\n",
    "                                                \n",
    "            bounds.append(\n",
    "                ( truevalue,\n",
    "                  data,\n",
    "                  [ \n",
    "                      ci[0], 'n/a', ci[1]\n",
    "                  ]\n",
    "                )  \n",
    "            )\n",
    "            \n",
    "        uppers = [ d[2][2] for d in bounds ]\n",
    "        lowers = [ d[2][0] for d in bounds ]\n",
    "        upperbounded = [ 1 if d[0] <= d[2][2] else 0 for d in bounds ]\n",
    "        lowerbounded = [ 1 if d[2][0] <= d[0] else 0 for d in bounds ]\n",
    "        covered = [ 1 if u*l > 0 else 0 for (u, l) in zip(upperbounded, lowerbounded) ]\n",
    "        width = [ d[2][2] - d[2][0] for d in bounds ]    \n",
    "        yield (ndata,\n",
    "                { \n",
    "                    'cov': numpy.mean(covered),\n",
    "                    'covstd': numpy.std(covered, ddof=1) / numpy.sqrt(len(covered) - 1),\n",
    "                    'ubcov': numpy.mean(upperbounded),\n",
    "                    'lbcov': numpy.mean(lowerbounded),\n",
    "                    'width': numpy.mean(width),\n",
    "                    'widthstd': numpy.std(width, ddof=1) / numpy.sqrt(len(width) - 1),\n",
    "                    'widthlo': numpy.quantile(width, q=[0.05])[0],\n",
    "                    'widthhi': numpy.quantile(width, q=[0.95])[0],\n",
    "                    'ub': numpy.mean(uppers),\n",
    "                    'lb': numpy.mean(lowers),\n",
    "                },\n",
    "              )\n",
    "\n",
    "def getenv():\n",
    "    import environments.DoubleDouble \n",
    "    wsupport = [ 0, 2, 1000 ]\n",
    "    env = environments.DoubleDouble.DoubleDouble(numactions=5, seed=45, wsupport=wsupport, expwsq=100)\n",
    "    return env\n",
    "    \n",
    "allres = []\n",
    "for (name, method) in [ \n",
    "                        ('MLE CI CV', lambda data, **kwargs: MLE.MLE.asymptoticconfidenceintervalwithcv(datagen=lambda: data, **kwargs)[0]),\n",
    "                        ('MLE CI', lambda data, rangefn, **kwargs: MLE.MLE.asymptoticconfidenceinterval(datagen=lambda: [(c,w,r) for c, w, r, _ in data], wmin=rangefn('wmin'), wmax=rangefn('wmax'), **kwargs)[0]),\n",
    "                        ('Asymptotic Gaussian', lambda data, rangefn, alpha, **kwargs: CITest.AsymptoticGaussian.interval([(c,w,r) for c, w, r, _ in data], rangefn('wmax'), alpha)),\n",
    "                        ('Binomial CI', lambda data, rangefn, alpha, **kwargs: CITest.ClopperPearson.interval([(c,w,r) for c, w, r, _ in data], rangefn('wmax'), alpha)),\n",
    "                      ]:\n",
    "    print('****** {} ******'.format(name))\n",
    "    res = []\n",
    "    for zzz in produceresults(getenv(), method, ndataperpt=50000, numpts=20):\n",
    "        res.append(zzz)\n",
    "        print('{}'.format(( zzz[0], { y: round(z, 4) for y, z in zzz[1].items() })), flush=True)\n",
    "    wmax = getenv().range('wmax')\n",
    "    allres.append((name, [(x[0] / wmax, x[1]) for x in res]))\n",
    "    del wmax\n",
    "    import pickle\n",
    "    pickle.dump( allres, open( \"epsilongreedy_ciwithcv_allres_big.p\", \"wb\" ) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "code_folding": [
     2,
     5
    ]
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "class FlassPlot:\n",
    "    from matplotlib import scale as mscale\n",
    "    \n",
    "    class ExpScale(mscale.ScaleBase):\n",
    "        from matplotlib import transforms as mtransforms\n",
    "\n",
    "        \"\"\"\n",
    "        Scales data in range -Infinity, 1 using log(1 - x)\n",
    "        \"\"\"\n",
    "\n",
    "        # The scale class must have a member ``name`` that defines the\n",
    "        # string used to select the scale.  For example,\n",
    "        # ``gca().set_yscale(\"mercator\")`` would be used to select this\n",
    "        # scale.\n",
    "        name = 'exp'\n",
    "\n",
    "        def __init__(self, axis, **kwargs):\n",
    "            \"\"\"\n",
    "            Any keyword arguments passed to ``set_xscale`` and\n",
    "            ``set_yscale`` will be passed along to the scale's\n",
    "            constructor.\n",
    "            \"\"\"\n",
    "            from matplotlib import scale as mscale\n",
    "            myscale = kwargs.pop(\"expscale\", 1.0)\n",
    "            mscale.ScaleBase.__init__(self, axis, **kwargs)\n",
    "            self.myscale = myscale\n",
    "\n",
    "        def get_transform(self):\n",
    "            \"\"\"\n",
    "            Override this method to return a new instance that does the\n",
    "            actual transformation of the data.\n",
    "            \"\"\"\n",
    "            return FlassPlot.ExpScale.ExpTransform(self.myscale)\n",
    "\n",
    "        def set_default_locators_and_formatters(self, axis):\n",
    "            \"\"\"\n",
    "            Override to set up the locators and formatters to use with the\n",
    "            scale.  This is only required if the scale requires custom\n",
    "            locators and formatters.  Writing custom locators and\n",
    "            formatters is rather outside the scope of this example, but\n",
    "            there are many helpful examples in ``ticker.py``.\n",
    "\n",
    "            In our case, the Mercator example uses a fixed locator from\n",
    "            -90 to 90 degrees and a custom formatter class to put convert\n",
    "            the radians to degrees and put a degree symbol after the\n",
    "            value::\n",
    "            \"\"\"\n",
    "\n",
    "            return axis\n",
    "\n",
    "        def limit_range_for_scale(self, vmin, vmax, minpos):\n",
    "            \"\"\"\n",
    "            Override to limit the bounds of the axis to the domain of the\n",
    "            transform.\n",
    "            \"\"\"\n",
    "            return vmin, vmax\n",
    "\n",
    "        class ExpTransform(mtransforms.Transform):\n",
    "            # There are two value members that must be defined.\n",
    "            # ``input_dims`` and ``output_dims`` specify number of input\n",
    "            # dimensions and output dimensions to the transformation.\n",
    "            # These are used by the transformation framework to do some\n",
    "            # error checking and prevent incompatible transformations from\n",
    "            # being connected together.  When defining transforms for a\n",
    "            # scale, which are, by definition, separable and have only one\n",
    "            # dimension, these members should always be set to 1.\n",
    "            input_dims = 1\n",
    "            output_dims = 1\n",
    "            is_separable = True\n",
    "\n",
    "            def __init__(self, scale):\n",
    "                from matplotlib import transforms as mtransforms\n",
    "                \n",
    "                mtransforms.Transform.__init__(self)\n",
    "                self.scale = scale\n",
    "\n",
    "            def transform_non_affine(self, a):\n",
    "                \"\"\"\n",
    "                This transform takes an Nx1 ``numpy`` array and returns a\n",
    "                transformed copy.  Since the range of the Mercator scale\n",
    "                is limited by the user-specified threshold, the input\n",
    "                array must be masked to contain only valid values.\n",
    "                ``matplotlib`` will handle masked arrays and remove the\n",
    "                out-of-range data from the plot.  Importantly, the\n",
    "                ``transform`` method *must* return an array that is the\n",
    "                same shape as the input array, since these values need to\n",
    "                remain synchronized with values in the other dimension.\n",
    "                \"\"\"\n",
    "                import numpy as np\n",
    "                \n",
    "                return np.exp(self.scale * a)\n",
    "\n",
    "            def inverted(self):\n",
    "                \"\"\"\n",
    "                Override this method so matplotlib knows how to get the\n",
    "                inverse transform for this transform.\n",
    "                \"\"\"\n",
    "                return FlassPlot.ExpScale.InvertedExpTransform(self.scale)\n",
    "\n",
    "        class InvertedExpTransform(mtransforms.Transform):\n",
    "            input_dims = 1\n",
    "            output_dims = 1\n",
    "            is_separable = True\n",
    "\n",
    "            def __init__(self, scale):\n",
    "                from matplotlib import transforms as mtransforms\n",
    "                \n",
    "                mtransforms.Transform.__init__(self)\n",
    "                self.scale = scale\n",
    "\n",
    "            def transform_non_affine(self, a):\n",
    "                import numpy as np\n",
    "                \n",
    "                return np.log(a / self.scale)\n",
    "\n",
    "            def inverted(self):\n",
    "                return FlassPlot.ExpScale.ExpTransform(self.scale)\n",
    "\n",
    "    @staticmethod\n",
    "    def yo():\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.figure(figsize=(4,2))\n",
    "        \n",
    "    @staticmethod\n",
    "    def forpaper():\n",
    "        import matplotlib.pyplot as plt\n",
    "        from matplotlib import scale as mscale\n",
    "\n",
    "        mscale.register_scale(FlassPlot.ExpScale)\n",
    "\n",
    "        SMALL_SIZE = 16\n",
    "        MEDIUM_SIZE = 22\n",
    "        BIGGER_SIZE = 24\n",
    "\n",
    "        plt.rc('font', size=SMALL_SIZE)          # controls default text sizes\n",
    "        plt.rc('axes', titlesize=SMALL_SIZE)     # fontsize of the axes title\n",
    "        plt.rc('axes', labelsize=MEDIUM_SIZE)    # fontsize of the x and y labels\n",
    "        plt.rc('xtick', labelsize=SMALL_SIZE)    # fontsize of the tick labels\n",
    "        plt.rc('ytick', labelsize=SMALL_SIZE)    # fontsize of the tick labels\n",
    "        plt.rc('legend', fontsize=SMALL_SIZE)    # legend fontsize\n",
    "        plt.rc('figure', titlesize=BIGGER_SIZE)  # fontsize of the figure title\n",
    "        \n",
    "    @staticmethod\n",
    "    def exppic(x, y, label, expscale, **kwargs):\n",
    "        import matplotlib.pyplot as plt\n",
    "        import matplotlib.ticker as ticker\n",
    "        plt.ylim(0.0, 1.01)\n",
    "        plt.plot(x, y, label=label, **kwargs)\n",
    "        plt.gca().set_xscale('log')\n",
    "        plt.gca().set_yscale('exp', expscale=expscale)\n",
    "        plt.ylim(0.0, 1.01)\n",
    "        plt.gca().yaxis.set_major_locator(ticker.FixedLocator([0.38,0.68,0.86,0.95,1]))\n",
    "\n",
    "        plt.legend()\n",
    "        \n",
    "    @staticmethod\n",
    "    def pic(x, y, label): \n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.semilogx(x, y, label=label)\n",
    "\n",
    "        #plt.loglog(x, y, label=label)\n",
    "        plt.legend()\n",
    "        \n",
    "    @staticmethod\n",
    "    def axeslabel(xlabel, ylabel):  \n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.xlabel(xlabel)\n",
    "        plt.ylabel(ylabel)\n",
    "        \n",
    "    @staticmethod\n",
    "    def title(title):\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.title(title)\n",
    "        \n",
    "    @staticmethod\n",
    "    def legend(**kwargs):\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.legend(**kwargs)\n",
    "        \n",
    "    @staticmethod\n",
    "    def savefig(filename):\n",
    "        import matplotlib.pyplot as plt\n",
    "        plt.savefig(filename, bbox_inches='tight')\n",
    "        \n",
    "    @staticmethod\n",
    "    def plt():\n",
    "        import matplotlib.pyplot as plt\n",
    "        return plt\n",
    "\n",
    "import pickle\n",
    "del allres\n",
    "allres = pickle.load(open( \"epsilongreedy_ciwithcv_allres_big.p\", \"rb\" ) )\n",
    "\n",
    "renameit = { \"Binomial CI\": \"Binomial\", \"Asymptotic Gaussian\": \"Gaussian\", \"MLE CI\": \"MLE\", \"MLE CI CV\": \"MLE+cv\" }\n",
    "FlassPlot.forpaper()\n",
    "for name, res in reversed(allres):\n",
    "    x = [ x[0] for x in res ]\n",
    "    y = [ x[1]['width'] for x in res ]\n",
    "    ylo = [ x[1]['widthlo'] for x in res ]\n",
    "    yhi = [ x[1]['widthhi'] for x in res ]\n",
    "    #ylo = [ x[1]['width'] - 1.96 * x[1]['widthstd'] for x in res ]\n",
    "    #yhi = [ x[1]['width'] + 1.96 * x[1]['widthstd'] for x in res ]\n",
    "    FlassPlot.plt().semilogx(x, y, label=renameit.get(name, name))\n",
    "    FlassPlot.plt().fill_between(x, ylo, yhi, alpha=0.15)\n",
    "\n",
    "FlassPlot.legend(loc=\"center\", bbox_to_anchor=[0.8,0.825])\n",
    "FlassPlot.axeslabel('n / $w_{max}$', 'CI width')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import pickle\n",
    "del allres\n",
    "allres = pickle.load(open( \"epsilongreedy_ciwithcv_allres_big.p\", \"rb\" ) )\n",
    "\n",
    "renameit = { \"Binomial CI\": \"Binomial\", \"Asymptotic Gaussian\": \"Gaussian\", \"MLE CI\": \"MLE\", \"MLE CI CV\": \"MLE+cv\" }\n",
    "FlassPlot.forpaper()\n",
    "for name, res in reversed(allres):\n",
    "    x = [ x[0] for x in res ]\n",
    "    y = [ x[1]['cov'] for x in res ]\n",
    "    ylo = [ x[1]['cov'] - 1.96 * x[1]['covstd'] for x in res ]\n",
    "    yhi = [ x[1]['cov'] + 1.96 * x[1]['covstd'] for x in res ]\n",
    "    FlassPlot.exppic(x, y, label=renameit[name], expscale=7.0)\n",
    "    FlassPlot.plt().fill_between(x, ylo, yhi, alpha=0.25)\n",
    "FlassPlot.exppic([ x[0] for x in res ], [ 0.95 for x in res ], label='_nolegend_', expscale=7.0, ls='dashed')\n",
    "FlassPlot.legend(loc=\"center\", bbox_to_anchor=[0.5,0.425])\n",
    "FlassPlot.axeslabel('n / $w_{max}$', 'empirical coverage')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}