[R] multiple comparison of interaction of ANCOVA

Jinsong Zhao jszhao at yeah.net
Tue Dec 13 09:37:19 CET 2011


Thank you very much for the suggestions.

On 2011-12-13 3:20, Richard M. Heiberger wrote:
> ...
>
> On Mon, Dec 12, 2011 at 7:00 AM, Jinsong Zhao <jszhao at yeah.net
> <mailto:jszhao at yeah.net>> wrote:
>
>     Apart from the data set here, is there a way to do multiple
>     comparison on the interaction of one way analysis of covariance?
>
> Interaction in ancova usually means the slopes are different.
> ...
>
<snip>
>
> A better solution is to use glht in the multcomp package instead of
> TukeyHSD.
> Then the statement is
>      confint(glht(mydata.aov, linfct=mcp(Trt="Tukey")))
> glht is better because it does not have the limitations of TukeyHSD.
> TukeyHSD is not applicable in your example because TukeyHSD doesn't work
> with covariates.
> The documentation ?TukeyHSD says "non-factor terms these will be dropped
> with a warning".
> With glht and the HH package loaded, you can get a better plot.
>      glht.mmc(mydata.aov, linfct=mcp(Trt="Tukey"))
>      plot(glht.mmc(mydata.aov, linfct=mcp(Trt="Tukey")))
> With the MMC plot ?HH:::MMC, "Each contrast is plotted at a height which
> is the weighted
> average of the means being compared"

Regards,
Jinsong



More information about the R-help mailing list