# Release 3.4 of the Real Statistics Resource Pack

The Real Statistics Examples Workbook has been updated to reflect the new release. You can also download this file for free (Download Examples). The website is in the process of being updated to reflect the new features. These changes will be made over the course of the next few days.

New features include:

New Non-parametric Tests data analysis tool

Provides access to non-parametric tests that were not previously accessible via the T Tests and Non-parametric Equivalents and ANOVA data analysis tools. Tests include:

• Friedman’s Test
• Runs Test
• Two Sample Komogorov-Smirnov Test
• Cochran’s Q Test
• Moods’ Test

New ROC and Classification Table data analysis tool

These capabilities are similar to those provided with the Logistic Regression data analysis tool, but don’t have to be used with logistic regression. Some new fetaures include:

• Two ways of calculating the AUC (area under the curve): rectangular and trapazoidal approximations
• Two data input formats: two columns (cutoff is via a row number) and three columns (cutoff is via a cutoff amount)

Power and Sample Size Requirements for Multiple Regression

A new option has been added to the Statistical Power and Sample Size data analysis tool to support multiple linear regression. In addition the following two new functions have been added:

REG_POWER(f, n, k, t, alpha, m) = statistical power of multiple regression test with sample size n and k independent variables for effect size f. Three measures of effect size are supported (based on argument t): if t = 1, then f = Cohen’s f-square, if t = 2 then f = R-square and if t = 0 then f = noncentrality parameter. As usual m = upper limit of the infinite sum (usually you can simply use the default)

REG_SIZE(f, k, pow, t, alpha, m) = minimim sample size required to achieve power pow where the other arguments are as for REG_POWER (except that t = 1 ot 2 only)

New Durbin-Watson data analysis tool

The Durbin-Watson test is now available as an option in the Linear Regression data analysis tool

Cochran’s Q Test functions

In addition to the data analysis tool described above, the following three new functions are now available:

• COCHRAN(R1, raw) = Cochran’s Q statistic
• QTEST(R1, raw) = p-value for Cochran’s Q test
• QRATIO(R1, raw) – returns a row range with the percentages of independent variables for Q test

Two data input formats are supported: raw data (raw = TRUE) and summarized data (i.e. a multi-variable frequency table)

Multi-row and multi-column sum functions

The following two new array functions hve been added:

• SUMROW(R1): outputs a column range with the counts of each of the rows in R1
• SUMCOL(R1): outputs a rowrange with the means of each of the columns in R1

Bug fixes

• A format error in the headings for Residuals and Cook’s D option of the Linear Regression data analysis tool has been corrected