search

gnu-octave / statistics

The Statistics package for GNU Octave
GNU General Public License v3.0
24 stars 22 forks source link

added equal_var argument to anova1 for Welch's ANOVA #15

Closed acp29 closed 2 years ago

pr0m1th3as commented 2 years ago

Nice additional feature in anova1 !!

acp29 commented 2 years ago

Thanks! I already wrote a welch's ANOVA function and distributed it on Mathworks file exchange (https://uk.mathworks.com/matlabcentral/fileexchange/61661-wanova?s_tid=srchtitle) so I just used some of that code to expand the capabilities of the anova1 function. I like that now anova1 and anova2 both have some functionality (i.e. Welch's test and nested ANOVA respectively) beyond what anovan is capable of (thanks to the simpler nature of the task in question)- they are not completely redundant with anovan. I have a bunch of other stats functions I wrote years ago that may be of interest too (in some form)... https://uk.mathworks.com/matlabcentral/fileexchange/61660-epsilon https://uk.mathworks.com/matlabcentral/fileexchange/61659-multicmp https://uk.mathworks.com/matlabcentral/fileexchange/72399-tiedrank2

This older version of multicmp has Shaffer step-down Bonferroni procedure which we could include in multcompare as a more powerful way of controlling Type 1 error in multiple pairwise comparisons: https://uk.mathworks.com/matlabcentral/mlc-downloads/downloads/submissions/61659/versions/3/download/zip

pr0m1th3as commented 2 years ago

Once I get the multcompare function ready for testing I will push it to main so you can include Shaffer step-down Bonferroni procedure. I won't be before 15 of September because work caught up with me. I think we can skip the extra multicmp function if we can add some useful extra functionality and keep matlab compatibility intact. epsilon would be a nice addition to have in the statistics package, but it would be ideally implemented with a more helpful or explanatory help file, some demos and BISTs of course. I am not sure what the tiedrank2 does differently in contrast to tiedrank which is already implemented.