v0.2.9

Release notes:

Dimensionality reduction

• transform() : fix for input data whose index is not a contiguous range (e.g. the output of remove_outliers()). Numerical and categorical components were joined on different indexes, which added rows with missing values. Both are now aligned on the input index, and an error is raised if the number of output rows differs from the input

Clustering

• Input index is kept when normalize=True
• describe_clusters() : fix for weighted statistics (wmean, wstd) when the index is not a contiguous range
• describe_clusters_cat() : observations are matched by position, as documented, also when using weights

Classifier

• plot_confusion_matrix() : cell percentages are now computed over the total number of observations only (they previously included the appended recall, precision and accuracy values)
• plot_confusion_matrix() : recall, precision and accuracy cells are now shown in grey

Other

• reset_index() is no longer needed after remove_outliers(); removed from the README example and replication script
• Added regression tests