new function `fitlm` and various minor bug fixes in `anovan`

gnu-octave / statistics

The Statistics package for GNU Octave

GNU General Public License v3.0

24 stars 22 forks source link

new function `fitlm` and various minor bug fixes in `anovan` #45

Closed acp29 closed 1 year ago

acp29 commented 1 year ago

anovan:

added support for missing data in nan (by ommission)
minor bug fix to prevent warning under some circumstances when creating STATS.coeffnames
fixed bug that occurred when trying to run model 'interaction' when there is only one factor in GROUP
changed it so that continuous variables are centered for all contrasts except "treatment"

fitlm:

new function for fitting linear models including demos and BISTs

multcompare

added support for fitlm

acp29 commented 1 year ago

but @pr0m1th3as the trace back tells you which function had the error...

[intercept, slope]=anovan(ones(10,1),2) error: GROUP must be a matrix with the same number of rows as Y error: called from anovan at line 335 column 9

I just took it out because I noticed that it was redundant, although I realise I forgot to do it for multcompare

I'm guessing a few other functions don't do this either:

[intercept, slope]=logistic_regression() error: 'y' undefined near line 107, column 107 error: called from logistic_regression at line 107 column 5

pr0m1th3as commented 1 year ago

Most functions do it this way because it is also consistent with how Matlab prints the error. I think it's best to keep it consistent, despite certain statistics functions might not be fully compliant yet.

pr0m1th3as commented 1 year ago

Thanks for this (the error style thing) and the new function as well.