Merge branch 'OperationChanges'

Operations/DN_RemovePoints.m

@@ -53,7 +53,7 @@
 %% Preliminaries
 % ------------------------------------------------------------------------------
 N = length(y); % time-series length
-doPlot = 0; % plot output
+doPlot = false; % plot output
 
 % ------------------------------------------------------------------------------
 %% Check inputs
@@ -108,6 +108,8 @@
 %% Compute output statistics
 % ------------------------------------------------------------------------------
 out.fzcacrat = CO_FirstZero(y_trim,'ac')/CO_FirstZero(y,'ac');
+out.ac1rat = acf_y_trim(1)/acf_y(1); % includes the sign
+out.ac1diff = abs(acf_y_trim(1)-acf_y(1));
 out.ac2rat = acf_y_trim(2)/acf_y(2); % includes the sign
 out.ac2diff = abs(acf_y_trim(2)-acf_y(2));
 out.ac3rat = acf_y_trim(3)/acf_y(3); % includes the sign

Operations/MF_CompareTestSets.m

@@ -236,7 +236,7 @@
 
     % Get statistics on output time series
     meandiffs(i) = abs(mean(yp.y) - mean(yTest.y));
-    stdrats(i) = abs(std(yp.y)/std(yTest.y));
+    stdrats(i) = std(yp.y)/std(yTest.y);
 
 %     % 1) Get statistics on residuals
 %     residout = MF_ResidualAnalysis(mresiduals);

Operations/NW_VisibilityGraph.m

@@ -138,6 +138,7 @@
 out.mink = min(k); % minimum degree
 out.rangek = range(k); % range of degree distribution
 out.iqrk = iqr(k); % interquartile range of degree distribution
+out.skewnessk = skewness(k); % skewness of degree distribution
 out.maxonmedian = max(k)/median(k); % max on median (indicator of outlier)
 out.ol90 = mean(k(k>=quantile(k,0.05) & k<=quantile(k,0.95)))/mean(k);
 out.olu90 = (mean(k(k>=quantile(k,0.95)))-mean(k))/std(k); % top 5% of points are
@@ -227,8 +228,8 @@
 
 % Extreme Value Distribution
 paramhat = evfit(k);
-out.evparm1 = paramhat(1);
-out.evparm2 = paramhat(2);
+out.evparam1 = paramhat(1);
+out.evparam2 = paramhat(2);
 out.evnlogL = evlike(paramhat,k);
 
 % ------------------------------------------------------------------------------

Operations/ST_LocalExtrema.m

@@ -63,7 +63,7 @@
     end
 end
 
-doPlot = 0; % plot outputs to a figure
+doPlot = false; % plot outputs to a figure
 
 N = length(y); % length of time series
 
@@ -92,20 +92,21 @@
 %% Buffer the time series
 % ------------------------------------------------------------------------------
 y_buff = buffer(y,wl); % no overlap
-% each *column* is a window of samples
+% Each *column* is a window of samples:
 if y_buff(end) == 0
     y_buff = y_buff(:,1:end-1); % remove last window if zero-padded
 end
-nw = size(y_buff,2); % number of windows
+numWindows = size(y_buff,2); % number of windows
 
 % ------------------------------------------------------------------------------
 %% Find local extrema
 % ------------------------------------------------------------------------------
 locmax = max(y_buff); % summary of local maxima
 locmin = min(y_buff); % summary of local minima
-abslocmin = abs(locmin); % absoluate value of local minima
-exti = find(abslocmin>locmax);
-locext = locmax; locext(exti) = locmin(exti); % local extrema (furthest from mean; either maxs or mins)
+abslocmin = abs(locmin); % absolute value of local minima
+exti = find(abslocmin > locmax);
+locext = locmax;
+locext(exti) = locmin(exti); % local extrema (furthest from mean; either maxs or mins)
 abslocext = abs(locext); % the magnitude of the most extreme events in each window
 
 if doPlot
@@ -136,8 +137,8 @@
 out.zcext = ST_SimpleStats(locext,'zcross');
 out.meanabsext = mean(abslocext);
 out.medianabsext = median(abslocext);
-out.diffmaxabsmin = sum(abs(locmax-abslocmin))/nw;
-out.uord = sum(sign(locext))/nw; % whether extreme events are more up or down
+out.diffmaxabsmin = sum(abs(locmax-abslocmin))/numWindows;
+out.uord = sum(sign(locext))/numWindows; % whether extreme events are more up or down
 out.maxmaxmed = max(locmax)/median(locmax);
 out.minminmed = min(locmin)/median(locmin);
 out.maxabsext = max(abslocext)/median(abslocext);

PeripheryFunctions/TS_local_clear_remove.m

@@ -1,10 +1,12 @@
-function TS_local_clear_remove(tsOrOps,idRange,doRemove,whatData)
+function TS_local_clear_remove(whatData,tsOrOps,idRange,doRemove)
 % TS_local_clear_remove     Clear or remove data from an hctsa dataset
 %
-% 'Clear' means clearing any calculations performed about a given time series
-% or operation, but keeping it in the dataset.
-% 'Remove' means removing the time series or operation from the dataset completely.
-% The result is saved back to the hctsa .mat datafile provided.
+% 'clear' (doRemove = false) means clearing any calculations performed about a
+% given time series or operation, but keeping it in the dataset.
+% 'remove' (doRemove = true) means removing the time series or operation from
+% the dataset completely.
+%
+% The result is saved back to the hctsa-formatted .mat data file provided as whatData.
 %
 %---INPUTS:
 % tsOrOps -- either 'ts' or 'ops' for whether to work with either time series
@@ -17,10 +19,15 @@ function TS_local_clear_remove(tsOrOps,idRange,doRemove,whatData)
 %---EXAMPLE USAGE:
 % This clears the data about the time series with IDs 1,2,3,4, and 5 from the hctsa dataset
 % stored in HCTSA.mat:
-% >> TS_local_clear_remove('ts',1:5,0,'HCTSA.mat');
+% >> TS_local_clear_remove('HCTSA.mat','ts',1:5,false);
 %
 % This *removes* the time series with IDs from 1:5 from the dataset completely:
-% >> TS_local_clear_remove('ts',1:5,1,'HCTSA.mat');
+% >> TS_local_clear_remove('HCTSA.mat','ts',1:5,true);
+%
+% IDs for a given keyword can be retrieved using TS_getIDs. This example removes
+% all time series from HCTSA.mat that have the keyword 'noise':
+% >> noiseIDs = TS_getIDs('noise','HCTSA.mat','ts');
+% >> TS_local_clear_remove('HCTSA.mat','ts',noiseIDs,true);
 
 % ------------------------------------------------------------------------------
 % Copyright (C) 2018, Ben D. Fulcher <ben.d.fulcher@gmail.com>,
@@ -48,7 +55,11 @@ function TS_local_clear_remove(tsOrOps,idRange,doRemove,whatData)
 %% Preliminaries and input checking
 %-------------------------------------------------------------------------------
 
-if nargin < 1
+if nargin < 1 || isempty(whatData)
+    whatData = 'raw'; % normally want to clear data from the local store
+end
+
+if nargin < 2
 	tsOrOps = 'ts';
 end
 switch tsOrOps
@@ -63,18 +74,14 @@ function TS_local_clear_remove(tsOrOps,idRange,doRemove,whatData)
 end
 
 % Must specify a set of time series
-if nargin < 2 || min(size(idRange)) ~= 1
+if nargin < 3 || min(size(idRange)) ~= 1
 	error('Specify a range of IDs');
 end
 
-if nargin < 3 % doRemove
+if nargin < 4 % doRemove
     error('You must specify whether to remove the %s or just clear their data results',theWhat)
 end
 
-if nargin < 4
-    whatData = 'raw'; % normally want to clear data from the local store
-end
-
 % ------------------------------------------------------------------------------
 %% Load data
 % ------------------------------------------------------------------------------

Toolboxes/Michael_Small/MS_complexity.m

@@ -1,9 +1,8 @@
-% cmp = MS_complexity(x,n);
-%
-% calculate the Lempel-Ziv complexity of the n-bit encoding of x. 
+function cmp = MS_complexity(x,n,binHow)
+% Calculate the Lempel-Ziv complexity of the n-bit encoding of x.
 %
 % cmp is the normalised complexity, that is the number of distinct
-% symbol sequences in x, divided by the expected number of distinct 
+% symbol sequences in x, divided by the expected number of distinct
 % symbols for a noise sequence.
 %
 % Algorithm is implemented in complexitybs.c
@@ -16,45 +15,49 @@
 % Series A, vol. 52. World Scientific, 2005. (ISBN 981-256-117-X) and the
 % references therein.
 
-function cmp = MS_complexity(x,n);
-
 if nargin < 2
     n = 2;
 end
+if nargin < 3
+    binHow = 'equiprobable';
+end
+%-------------------------------------------------------------------------------
 
 if length(n) > 1
-   for ni = 1:length(n)
-        cmp(ni) = MS_complexity(x,n(ni)); 
-   end
-else,
-
-    if 1,
-        %do the binning, with equi-probably bins
+    % Can run with multiple values of the number of symbols, n:
+    for ni = 1:length(n)
+        cmp(ni) = MS_complexity(x,n(ni),binHow);
+    end
+else
+
+    % do the binning, with equiprobable bins
+    switch binHow
+    case 'equiprobable'
         x=x(:);
         nx=length(x);
         [xn,xi]=sort(x+eps*randn(size(x))); %introduce randomness for ties
         y=zeros(nx,1);
         y=1:nx;
         y=floor(y.*(n/(nx+1)));
         x(xi)=y;
-    else,
-        %do binning with equal width bins
+    case 'equiwidth'
+    % else,
+    %     %do binning with equal width bins
         x=x(:);
         nx=length(x);
         minx=min(x);
         maxx=max(x);
         stepx=(maxx-minx)/n;
         y=zeros(nx,1);
-        while minx<maxx,
+        while minx<maxx
             minx=minx+stepx;
             y=y+double(x<minx);
-        end;
+        end
         x=floor(y);
-    end;
-
-    %compute complexity with complexitybs
-    cmp=MS_complexitybs(x);
+    end
 
-end;
+    %compute complexity with complexitybs
+    cmp = MS_complexitybs(x);
+end
 
-end
+end