Hi Rose!

First of all, many thanks for your reply! :)

I've been reading in forums, and if I understood well, there should be no differences between:

aov_results

I recommend that you treat month as a continuous predictor and use lmer in R - it’s in the lme4 package. By treating month as a continuous variable you avoid the numerous post hocs.

If you have questions, send me a message or even your data. Your analysis would be very simple to run.

Ask yourself these two questions: "Do I want to know the exact time point when there is a "significant difference" or do I want to know whether WT compared to KO gains more weight across time?"

The former question can be answered by a Repeated Measures ANOVA model. To add to Dr. Young's comment: The "at which time point(s) is this difference significant?" is a question you ask by comparing every time point individually. If you do that, you eat up degrees of freedom and therefore lose power in your analysis to find an effect.

The latter question, which is the one I think you actually want to know, is about whether or not the group's *Rate of Change* in weight is different across the whole time period. Dr. Young's suggestion will allow you to look at the thing I think you are more interested in: rate of change. By treating your Month variable as continuous instead of a factor, your interaction term will allow you to compute the slope of the line for each group and test if the slopes of the weight gain across time are different between WT and KO. In other words, you can test whether the rate of change in weight gain is significantly different between the two groups without having to run tons of post hocs AND you will have more statistical power.

Think of it this way. In a chemical reaction if I said there was "more" of chemical x present at time "t" instead of giving an amount or a function of the change in chemical x across time, I am giving much less information about the functional relationship to my reader. In the end, prediction, explanation, and control (the ultimate goals of science) depend on knowing functional relationships with both direction and magnitude. With the repeated measures ANOVA method, you only get direction, not magnitude or a functional relationship AND you reduce your power.

Unless your question really is about an exact time point that is theoretically important, I would use lme4::lmer. It is more user-friendly than lme and is still running a linear mixed effects regression:

library("lme4")

#adjust coding to Sum to Zero (contrast) coding so the interpretation is identical

#to ANOVA

contrasts(DATA_MALE$Genotype)=contr.sum(2)

contrasts(DATA_MALE$Genotype)

#center your continuous time variable to avoid variance inflation

DATA_MALE$c.month

Thanks a lot to all of you!, Specially to Michael E Young and Thomas Wukitsch for your time and efforts.😃

I've applied your recommendations to my dataset (now concerning the change in wound area in a cell culture, but very similar to the first posted problem...).

I have measured this area at 3 different timepoints in 4 different groups (celltypes: Mock, A1, FARP1 and A1K51R).

The model renders the following results. (For now, cTimepoint correponds to the Timepoint variable treated as continous, but without substracting the mean (for a major understanding)):

lmer_results lmer_results lmer_results