conovertest.  Conover-Iman test of multiple comparisons using rank sums
Program by Alexis Dinno.
Support: alexis.dinno@pdx.edu
Version: 1.1.6 (updated January 14, 2025)
Distribution-Date: 14jan2025

This package includes conovertest and the data set homecare which is 
reproduced from Conover & Iman's 1979 paper. conovertest performs 
approximate t tests of nonparametric null hypothesis of stochastic 
equality, and under certain assumptions, tests nonparametric null hypotheses
of median equality. Conover-Iman test is an appropriate multiple pairwise 
comparison procedure following the rejection of the null hypothesis in a
Kruskal-Wallis test.

This package is written for STATA versions 8 - 14, however, requests for a 
version compatible with versions of STATA earlier than v8 are untenable since 
I do not have access to the software.

All feature and bug requests are welcome and will be considered.

HISTORY

14jan2025 (1.1.6)  minor code refactor to reduce risk of possible (unlikely)  
                   namespace collisions with user variables.

03may2018 (1.1.5)  corrected an obscure bug where having variables in
                   memory starting with "P" could break conovertest.

28feb2018 (1.1.4)  corrected a bug where significant rejections were not 
                   underlined or starred in the output when using FDR 
                   options for ma().

26oct2017 (1.3.3)  added the altp option to express p-values as 
                   p = P(|T| >= |t|) in output, and include the rejection rule 
                   for Ho explicit in the output. documentation changes.

27apr2017 (1.1.2)  made conovertest play nice with Stata versions 8 through 13

08apr2017 (1.1.1)  corrected a bug where significant rejection was indicated
                   incorrectly for bonferroni and sidak multiple comparisons
                   adjustments, that was created in the previous update.

06apr2017 (1.1.0)  added a new list option, to include all pairwise
                   tests as a list. various minor improvements to output
                   made in general, including specific reporting of level
                   of significance, and indicating which tests are rejected
                   for no multiple comparisons adjustments, the Bonferroni
                   adjustment, and the Sidák adjustment.

08feb2017 (1.0.10) fixed a formatting bug that mangled the table at the end
                   of non-wrapping output with many comparisons.

06sep2016 (1.0.9)  made conovertest play nicely when groupvar is
                   specified as a string variable rather than a numeric 
                   variable with value labels.

07jan2016 (1.0.8)  fixed a fatal yet ridiculous bug that crept in.

06jan2016 (1.0.7)  fixed a bug that could sometimes hide some of the output 
                   of non-wrapping results when the number of groups was 
                   greater than 7.

16dec2015 (1.0.6)  updated to reflect new web domain.

27apr2015 (1.0.5)  fixed a bug where output was broken when the number of
                   groups was greater than 7 and the wrap option was not
                   used.

17apr2015 (1.0.4)  even better handling of labels, now featuring explicit
                   warnings and separate handling of incomplete labeling
                   of groupvar.

03mar2015 (1.0.3)  fixed yet another variation on the value label bug, where
                   values of the group variable not conforming to 1 – k
                   got munged. seemingly fixed now. until the next variation.

08feb2015 (1.0.2)  fixed two bugs: (1) the pairwise tests table was labeled
                   Row Mean-Col Mean, but actually reported Col Mean-Row Mean.
                   the signs of the test statistics were thus reversed. The
                   table is now labeled Col Mean-Row Mean, but there is a new
                   rmc option to report the opposite. (2) when using value
                   labels in conjunction with either if or in conditionals
                   the correct tests were reported, but with the wrong labels
                   attached. this bug did not affect results presented using
                   the nolabel option.

08jan2015 (1.0.1)  fixed a serious bug where conovertest did not properly
                   handle if and in statements

03dec2014 (1.0.0)  first release of the conovertest package
