[R] Multiple comparison on lme model with 2 fixed factors

Dieter Menne dieter.menne at menne-biomed.de
Tue Aug 18 10:18:11 CEST 2009

Billeke wrote:
> mult2 <- glht(lmelac.ex, linfct = mcp("ath:time" = "Tukey"))

I don't have the time to check, but got a mail for Torsten (author) with the
following idea:

glht(lmeglu, linfct = c("Ath1:tim2 - Ath1:tim3 = 0", ...))

He suggested to build the matrix explicitly instead; that's what I normally
do, but I admit that I use an Excel spreadsheet to get the coefficients
right in such a case.

Technically, using time as a categorical parameter is not recommended
because of serial correlations. Try to fit a linear or linearized model
instead to time, or, if you cannot do it otherwise, a nonlinear one.

I too often see my colleagues asking questions like "how many days after
treatment start do I first see significant effects on my lymphocyte count"?
This gives many p-values for a publication, and that's what medical
reviewers love. Your approach to use multcomp to correct for this is a fair
way out, but doing serious trend analysis is better.


View this message in context: http://www.nabble.com/Multiple-comparison-on-lme-model-with-2-fixed-factors-tp25002051p25020461.html
Sent from the R help mailing list archive at Nabble.com.

More information about the R-help mailing list