I met such problem. I have measured the respiration of animals in in 5 time intervals. Within each interval I have animals asigned to 4 treatments. In my main analysis I consider both: treatment and interval as fixed categorical factors. I obtained significant interaction. I'd like to check now post hoc with groups are differing. If I use classical post hoc tests I make many unnecessary comparisions (I have 20 groups in such scenario) between treatment from variuos intervals what decreases power of the test and does not interest me. I'd like to ask if that is any solution of that problem? I thought about Dunnett's test, but there I would compare everything with control group (what decreases number of possible outcomes and does not meet hypotheses I'd like to test). I'd like to compare treatments in one interval, but do not compare treatments between intervals. So far I do not know any procedure to do it. I also tried to find proper test to compare all groups pairwise but do not loose the power of the test. I've done already Tukey, Fisher (but than I learned that this test is rather of hictorical use) and Duncan tests. I read that quite good is Ryan's test (powerfull but roboust to type I error), but I cannot find any procedure to do that in R for lmer class objects (I have random factors as well in the model). I'd be very grateful for any help.

I met such problem. I have measured the respiration of animals in in 5 time intervals. Within each interval I have animals asigned to 4 treatments. In my main analysis I consider both: treatment and interval as fixed categorical factors. I obtained significant interaction. I'd like to check now post hoc which groups are differing. If I use classical post hoc tests I make many unnecessary comparisions (I have 20 groups in such scenario) between treatment from variuos intervals what decreases power of the test and does not interest me. I'd like to ask if that is any solution of that problem? I thought about Dunnett's test, but there I would compare everything with control group (what decreases number of possible outcomes and does not meet hypotheses I'd like to test). I'd like to compare treatments in one interval, but do not compare treatments between intervals. So far I do not know any procedure to do it. I also tried to find proper test to compare all groups pairwise but do not loose the power of the test. I've done already Tukey, Fisher (but than I learned that this test is rather of hictorical use) and Duncan tests. I read that quite good is Ryan's test (powerfull but roboust to type I error), but I cannot find any procedure to do that in R for lmer class objects (I have random factors as well in the model). I'd be very grateful for any help.