Version 0.8.0

Moved Exporting of logs and pickles to subdirectories
This required significant editing of the parallel processing code.

Version 0.7.1

There have been several updates in this version, relative to the last release.

Graphs are now exported to their own subdirectory.
Parameter names are no longer required -- CheKiPEUQ will autopopulate if they are not provided.
CheKiPEUQ will detect cases with single responses with multiple values in an array that need to be split as if they were separate responses to achieve evaluatable likelihood terms.
Turned on convertToSamples by default for doOptimizeNegLogP multistart

Version 0.6.7

The primary change in this version update is that responses with large numbers of values (threshold of 100 by default) have their likelihoods evaluated in a loop across variances rather than with a covariance matrix, leading to slower but linear scaling for larger data sets, such that thinning is not required.
The heuristic in the absence of uncertainties has been changed to continuous.

Version 0.6.5

Some CheKiPEUQ dependencies which were only used in specific functions are now only imported in those functions. This allows for more sparse CheKiPEUQ installations.

Version 0.6.4

Added weighting feature.

Additionally, fixed a minor bug in get loglikelihood for cases like 2 responses with 2 points (in those cases, there could be an error if the code treated the uncertainties like a square covariance matrix if they were not intended to be such)

Version 0.6.3

exportAllSimulatedOutputs is now working and tested and works with Metropolis Hastings as well multi-start.

Version 0.6.1

Various changes for responses uncertainties and multiple responses.
a) Applied 2dNested functions in more places
b) fixed an error in the getLogLikelihood logic for multiple responses
c) extended plotting of responses simulated uncertainties to accommodate multiple responses.
d) updated heuristic for observed uncertainties (when not provided) to be based on absolute value to accommodate negative response values.
e) extended logic to allow full covmat for each response (untested)

Version 0.4.9

This update makes improvements to responses observed uncertainties and simulated responses uncertainties, in particular for their compatibility with providing None types and functions rather than numbers, zeros, or blank lists. During this update, a bug in simulated responses uncertainties was found in DoE due to a 'typo' (the copying and pasting of the wrong variable).

Version 0.4.8

Bug fix for when a function is provided instead of an array-like object for simulated response uncertainties.

Version 0.4.7

Improved plotting: When there are many parameters, by default there will be no large scatter_matrix made nor a combined mumpce plot. Now, individual contour plots are made by default for each possible parameter pairing. Additionally, this update has improved Contour plot settings to have more 'contour_settings_custom' arguments, and the same for the scatter matrix.