I am pleased to announce Release 5.8 of the Real Statistics Resource Pack. The new release is now available for free download at Download Resource Pack for Excel 2007, 2010, 2013 and 2016 (Windows) environments. Release 5.8 will be available for Excel 2011 and Excel 2016 for Mac in about a week.
The examples workbooks have been updated for compatibility with the new release. In addition, the example workbooks have now been reorganized. There are now ten such workbooks as described below. Over the course of the next several days, the website will be updated for compatibility with the new release.
If you are getting value from the Real Statistics website or software, I would appreciate your donations to help offset the costs of the website by going to Please Donate.
The following is a summary of the new features in Release 5.8. The focus of this release is on improving the post-hoc tests following ANOVA, ANCOVA, Kruskal-Wallis and Friedman’s test.
Multiple Tests data analysis tool
A new Multiple Tests data analysis tool has been added to the Misc tab that supports the Bonferroni, Dunn-Sidàk, Holm, Hochberg, Benjamini-Hochberg and Benjamini-Yekutieli tests. These tests can be performed following multiple statistical tests in order to control familywise error.
One Factor ANOVA data analysis tool
The following new test options have been added to this data analysis tool:
- ANOVA follow-up: REGWQ test and Pairwise t tests
- Kruskal-Wallis follow-up: Conover test and Pairwise Mann-Whitney tests (both the versions based on the normal approximation and the exact test)
The format of most of the follow up tests have been revised. In particular, the p-value and other key statistics are now reported for all pairs of groups for the Tukey HSD, Games-Howell, REGWQ, Pairwise t, Nemenyi, Dunn, Schaich-Hamerle, Conover, Pairwise Mann-Whitney and Pairwise Mann-Whitney Exact tests.
For Contrasts, the number of tests is now explicitly specified (the Bonferroni or Dunn-Sidàk correction is applied based on this number).
The Nemenyi test has been revised to use the Rank-mean instead of Rank-sum to better handle groups of unequal size.
A new REGWQ function has been added to support the REGWQ post-hoc test.
One Repeated Measures Anova data analysis tool
The One Repeated Measures Anova data analysis tool has been revised as follows:
- The dialog box has been reformatted
- The Input Range can now optionally include column headings only. Row headings should not be included in the input range. This makes data entry easier when there are column headings but no row headings.
- A new modified version of the Tukey HSD test has been added which uses the pairwise standard errors. This is more useful than the other version which uses the combined standard error based on the MSE statistic.
- The p-value and other key statistics are now reported for all pairs of groups for both versions of the Tukey HSD test.
- A Pairwise t test post-hoc test has been added, which calculates the p-values of the paired sample t test for all pairs of groups.
Friedman Test data analysis tool
A new Friedman Test data analysis tool has been added to the Anova tab that supports the Friedman test and the Nemenyi, Conover and Pairwise signed-ranks tests (both the versions based on the normal approximation and the exact test).
For the Friedman Test a more accurate F test has been added in addition to the chi-square test commonly used.
Non-parametric Tests data analysis tool
Support for the Sign Test has been added to the Non-parametric Tests data analysis tool. The Friedman test has been removed from this tool since it is now supported by its own data analysis tool.
When the Sign Test option is selected, a paired sample test is performed when the data in the Input Range contains two columns and the one sample test is performed otherwise.
Other enhancements to the Analysis of Variance data analysis tools
The format of the dialog boxes for the Randomized Complete Block Design, Latin Squares, Split-Plot, Two Factor ANOVA Follow-up and ANCOVA data analysis tools has been changed. In particular, for Contrasts the number of tests is now explicitly specified (the Bonferroni or Dunn-Sidàk correction is applied based on this number).
In addition, the p-value and other key statistics are now reported for all pairs of groups for the Tukey HSD test (except for the Interaction effect following a Two Factor ANOVA, which has not been changed).
Other Enhancements
In addition to the SignTest(R1,med,tails) version of the sign test function, which supports a one sample test, a new version, SignTest(R1,R2,tails), has been added to support the paired sample sign test.
Nine new named colors have been added — crimson, tomato, khaki, lavender, violet, sky (blue), chocolate, saddle (brown) — bringing the total to 33 named colors. In the Color Assignment data analysis tool, these colors are now listed in alphabetical order.
A ties correction argument has been added to the KRUSKAL(R1, ties) and KTEST(R1, ties) functions. When ties = TRUE (default) then a ties correction is applied.
Ties correction and test type arguments have been added to the FRIEDMAN(R1, ties, chi) and FrTEST(R1, ties, chi) functions. When ties = TRUE (default) then a ties correction is applied. When chi = TRUE (default) then a chi-square test is used, while if chi = FALSE then a F test is used.
Previously when using the Box Plot with Outliers option of the Descriptive Statistics and Normality data analysis tool, elements more than 1.5 times Q3 + 1.5(Q3-Q1) or less than Q1 – 1.5(Q3-Q1) were flagged as outliers. Here 1.5 is the Outlier Multiplier factor. This value can now be set by the user (2.2 by default).
Examples Workbooks Reorganization
The following ten examples workbooks are now supported. In parentheses is a list of the Real Statistics website main menu topics covered in each examples workbook.
- Basics (Basics menu plus Mathematical Notation and Concepts, Excel Capabilities, Matrices and Iterative Procedures from the Appendix menu)
- Distributions (Distributions menu except for Testing for Normality and Symmetry)
- Non-parametric Tests (Non-parametric Tests from the Miscellaneous menu plus Testing for Normality and Symmetry from the Distributions menu)
- Correlation and Reliability (Correlation and Reliability from the Miscellaneous menu)
- Anova Part 1 (One-way ANOVA and Factorial ANOVA from the ANOVA menu)
- Anova Part 2 (Topics from the ANOVA menu not included in Anova Part 1)
- Regression Part 1 (Linear Regression and Multiple Regression from the Regression menu)
- Regression Part 2 (Topics from the Regression menu not included in Regression Part 1, plus Survival Analysis and Handling Missing Data from the Miscellaneous menu
- Multivariate Analysis (Multivariate menu plus Linear Algebra and Advanced Matrix Topics from the Appendix menu)
- Time Series Analysis (Time Series from the Miscellaneous menu plus Other Mathematical Topics from the Appendix menu)
Bug Fixes
- Fixed an error in the TiesCorrection function which didn’t count a number and its negative as tied in the Signed Ranks tests.
- Fixed the error in the MWDIST(x, n, tails) function reported on July 31. This error only occurred rarely, in particular when C(n+1, 2) was even and x = C(n+1,2)/2.
- Fixed error in the DCRIT and DPROB functions for values of α between .05 and .1
- Fixed error in the QCRIT and QPROB functions when k ≤ 20
- Fixed error in the Correlation data analysis tool, in which the correct output was displayed along with an erroneous error message.
- Corrected an error which occasionally produced a p-value larger than 1 from the SignTest function.