Scratch Notes

Scratch Notes

loading json

>>> import os
>>> import json
>>> import statsmodels.stats.tests
>>> cur_dir = statsmodels.stats.tests.__path__[0]
>>> fp = open(os.path.join(cur_dir, "results", "influence_lsdiag_R.json"))
>>> lsdiag = json.load(fp)
>>> fp.close()
>>> lsdiag.keys()
[u'cooks', u'std.res', u'dfits', u'std.err', u'std.dev', u'correlation', u'cov.unscaled', 
 u'stud.res', u'hat', u'cov.scaled']

used in statsmodels\stats\tests\test_diagnostic.py

example : export results from R using json

library(lmtest)
library(sandwich)
library(rjson)

options(digits=20)

mkhtest <- function(ht, name, distr="f") {
    cat(name); cat(" = dict("); 
    cat("statistic="); cat(ht$statistic); cat(", ");
    cat("pvalue="); cat(ht$p.value); cat(", ");
    cat("parameters=("); cat(ht$parameter, sep=","); cat(",), ");
    cat("distr='"); cat(distr); cat("'");
    cat(")");
    cat("\n\n")
}

d = read.csv("path_to\\statsmodels\\datasets\\macrodata\\macrodata.csv", header=TRUE)

names(d)

dinv <- diff(d$realinvs)
dgdp = diff(d$realgdp)
lint = d$realint[1:202]
tbilrate = d$tbilrate[1:202]

ginv <- 400 * diff(log(d$realinv))
ggdp <- 400 * diff(log(d$realgdp))

d2 = data.frame(ginv, ggdp, lint, tbilrate)

fm_level <- lm(realinv ~ realgdp + realint, data = d)

fm <- lm(ginv ~ ggdp + lint, data=d2)

nw = NeweyWest(fm, lag = 4, prewhite = FALSE, verbose=TRUE)
nw

nw2 = NeweyWest(fm, prewhite = FALSE, verbose=TRUE)
nw2

rt = raintest(fm)
mkhtest(rt, 'raintest', 'f')
jt = jtest(fm, lm(ginv ~ ggdp + tbilrate))
toJSON(jt)