GNOME Bugzilla – Bug 309497
General ANOVAs with or without repetition
Last modified: 2018-05-22 13:11:00 UTC
Distribution/Version: Gentoo I would like if gnumeric could perform an ANOVA test where both factors have replication. Something like: Test X Day, with two tests and two days, and many observations per day/test (like 1 per min, say). I believe it would need a "number of COLUMNS per sample" option, to perform it in a similar way gnumeric does the 2-factor ANOVA with repetition. Other programs (SPSS, StatView) use a collumn or more for coding of the treatments (Day1/Test1; Day1/Test2; Day2/Test1; Day2/Test2) while the rest of the row would hold the repeated measures (min1, min2, min3...).
Please explain this in more detail. Gnumeric can currently perform a 2 factor ANOVA with replicated observation, ie. something like "Test X Day, with two tests and two days, and many observations per day/test", it just assumes that the all replications are within for a given test/day combination are all within a certain row range within a fixed column. I can see a need to allow for one measurement per row with an indicator column for each factor, but that would still be differnet from the format you suggests where teh repated observation are all within the same row.
I have 162 min for each test, and my design is 2 tests (within group) and 2 days (between groups). The documentation in http://www.gnome.org/projects/gnumeric/doc/anova.html is unclear about how to perform it in Gnumeric. The indicator collumns would be a way of making true unlimited factorial analyses. The suggestion of "number of collumns per sample" was just a suggestion that took into account the present Gnumeric's way of handling the data. This suggestion was given as an alternative, and is unrelated to the "category collumn" approach. I actually would prefer an indicator collumn. I did not touch on the within/between groups issue on the original post, but yes, this would also have to be tackled somehow.
Adding up to what I've just say, a simple way of dealing with it would be to make the comparisons between collumns to be between groups, and within collumns, to be within groups. But of course, I have no way of knowing the easiest way through the source code... I would still feel more comfortable with the indicator column approach.
Created attachment 48761 [details] Sample data for the desired ANOVA comparison The file should be self-explainable. Row 1 contains captions for the columns. Column A has the subject's names, B contains a within-group factor, C contains a between-group factor, and the remaining ones are measures of mean blood pressure per minute. Each row is a different test. There are missing values as well.
Looking in that file I see 3 factors in a nested design.I agree that it would be nice if gnumeric could handle such an anova.
I should have changed thte title.
-- GitLab Migration Automatic Message -- This bug has been migrated to GNOME's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.gnome.org/GNOME/gnumeric/issues/39.