Release 3.0.4

• Fixing so that other multiple correction methods than Bonferroni are applied correctly.

Release 3.0.3

• Relaxing version requirements for scipy and pandas to allow versions 2.x
• Corrected typo in sample size calculator docstring

Release 3.0.2

• Added docstring to Experiment
• When using variance reduction and there is no pre-exposure data, so that the corresponding matrix is not invertable we now fall back to using a matrix with zeroes, leading to no variance reduction for this case, instead of crashing.

3.0.1

• Added reference level point estimate to the hover box of difference plots

Release 3.0.0

• Dropped support for python 3.6
• Improved formatting of some chart labels
• Made sure NIM lines in difference charts are always red when metrics are inferior

Release 2.7.7

• Fixed bug that led to unexpected behaviour when using non_inferiority_margins=False. Not passing False produces the same result as passing None
• Fixed bug in chartify grapher that caused a crash when attempting to plot a mix of metrics where only some had non-inferiority margins

Release 2.7.6

• Fixed bug in compute_sequential_adjusted_alpha where we since 2.7.4 were taking the max sample size rowwise

Release 2.7.5

• Major refactoring, splitting the code into more files for improved readability
• Fixed bugs related to group sequential testing that resulted in to narrow confidence bands for tests with multiple treatment groups
• Bump Chartify version
• Minor changes to get rid of warnings

Release 2.7.4

Fixed bug in sample size calculator check for binary metrics when there are nans

Release 2.7.3

• Fixed bug in SampleSizeCalculator.optimal_weights_and_sample_size.

• Added check to make point estimates and variances match for binary metrics when using the sample size calculator.