statistiXL 1.6 Released
- The Wilcoxon Paired-Sample Test has been completely revamped. statistiXL now allows the user to perform a standard Wilcoxon analysis (which effectively ignores tied values) and a more powerful analysis that compares the two data-sets taking these tied values into consideration. This new analysis is a closer approximation of the parametric paired t-Test and is a feature not found in many statistics packages. Due to the importance of this fix we would urge all users of statistiXL to upgrade to this new version.
- In previous versions of statistiXL the nonparametric Mann-Whitney U Test would fail for large sample sizes. This limitation has now been removed by introducing a normal approximation to the U statistic for such large samples.
- Previous versions of statistiXL suffered a performance hit when running on workbooks containing many formulae and with Excel's Automatic Calculation feature enabled. statistiXL's speed in this situation has now been greatly improved.
- Histograms are now possible with a much larger number of divisions (i.e. a greater number of columns).
- The Goodness of Fit tests now support much larger sample sizes.
- A bug within the Wilcoxon routine caused statistiXL to present incorrect P values when the paired datasets contained tied values (i.e. when the value for a given row was the same in each set). The Wilcoxon module has been completely rewritten to remove this bug and to expand the power of the analysis (see above). Due to the importance of this fix we would urge all users of statistiXL to upgrade to this new version.
- Multiple and partial correlation analyses now display the correct descriptive statistics. In previous versions, the descriptive statistics displayed were for the 'X Variables'. This has now been replaced with the proper 'Y Variable(s)' descriptives.
- Goodness-of-Fit tests no longer fail when expected values= 0.
- Lines are no longer drawn between adjacent box plots in Excel 2003.