I have data that is classified into 3 times (1-3) in each of 4 weeks (1-4) in a repeated measures design (5 subjects). (Total 3 times x 4 weeks = 12 mean values throughout the experiment)
I want to present lsmeans of the three times compared within each week.
For Bonferroni correction, which is correct; dividing p=0.05/3 = 0.0167 (three comparisons within a week; time1 vs. time2; time1 vs. time3; time2 vs. time3). Or, p=0.05/66 = 0.00076 (all possible comparison, which is not my aim).
If you plan on looking at only three comparisons you divide by 3. Best wishes, David Booth For a reference, see p. 57ff of Jason C. Hsu, Multiple Comparisons Theory and Methods, Chapman & Hall
In SAS, you can use the proc mixed to get the lsmeans. With the ODS facility you then can output these and run the proc multtest for the Bonferroni corrections.
See the SAS program in attachment.
If you do not have SAS, you can run the program with SAS Studio on the "SAS on demands" website: https://odamid.oda.sas.com/SASODAControlCenter/
If you only want to compare within the main effects separately, the run:
proc mixed data=b;
class pp times weeks;
model dv = times weeks times*weeks;
repeated / type=vc sub= pp;
lsmeans times / diff;
lsmeans weeks / diff;
ods output diffs=p_vals;
run;
proc print data=p_vals;
run;
data p_vals;
set p_vals;
rename probt=raw_p;
run;
proc multtest pdata=p_vals bon out=adjust;
run;
proc print data=adjust;
format bon_p pvalue7.4;
run;
In R:
https://www.researchgate.net/post/Could_someone_advise_me_on_post-hoc_analysis_in_R_with_repeated_measures_and_mixed_models
Or:
https://www.youtube.com/watch?v=wWjVQ8eqR9k
This is a continuation of the documentation below, but in R
In SPSS:
http://www.discoveringstatistics.com/docs/repeatedmeasures.pdf
- Badges
- Science topic