Hi:
See below.
On Mon, Aug 30, 2010 at 2:21 PM, Bruce Johnson
<Bruce.EJohnson@verizon.net>wrote:
> I am trying to do post-hoc tests associated with a repeated measures
> analysis with on factor nested within respondents.
>
> The factor (SOI) has 17 levels. The overall testing is working fine, but I
> can't seem to get the multiple comparisons to work.
>
> The first step is to "stack" the data.
>
> Then I used "lme" to specify and test the overall model.
>
> Finally I'm trying to use "glht" to do multiple comparisons.
>
> Here's my code and the resulting output.
>
> > ImpSoi<-ImpData[,c("iobs", soi)]
>
> > ImpSoi[1:5,]
>
> iobs soi1 soi2 soi3 soi4 soi5 soi6 soi7 soi8 soi9 soi10 soi11 soi12 soi13
> soi14 soi15 soi16
>
> 1 32 7.00 7.00 7.00 7.00 7.00 6.00 7.0 7.00 5.00 7.00 7.00 7.00 7.00
> 5.00 7.00 6.00
>
> 2 70 5.95 4.95 7.00 4.95 5.20 5.40 4.2 3.95 4.15 4.95 4.85 4.95 6.75
> 5.95 5.20 5.10
>
> 3 78 3.00 1.00 4.75 2.75 3.00 4.50 4.0 4.00 1.50 4.00 4.00 4.50 2.50
> 3.00 3.75 4.00
>
> 4 104 3.75 3.50 6.25 5.25 4.25 3.75 4.0 5.25 4.75 4.75 5.00 5.75 6.00
> 4.00 4.75 3.75
>
> 5 117 5.00 5.00 4.00 5.00 2.00 4.62 5.0 4.00 4.00 4.00 5.00 4.00 4.70
> 4.70 5.00 2.00
>
> soi17
>
> 1 7.00
>
> 2 5.15
>
> 3 4.00
>
> 4 5.50
>
> 5 4.00
>
> > stack <-
>
reshape(ImpSoi,varying=soi,timevar="solution",idvar="iobs",sep="",
> dir="long")
>
> > solution <- factor(solution,levels(1:17),)
>
> > stack[1:5,]
>
> iobs solution soi
>
> 32.1 32 1 7.00
>
> 70.1 70 1 5.95
>
> 78.1 78 1 3.00
>
> 104.1 104 1 3.75
>
> 117.1 117 1 5.00
>
Try str(stack) at this point. Is solution a factor? (Hint: You defined
solution *outside* of the stack data frame. Now look at ls(), which
tells you the variables in your global workspace.)
> > Lmes.mod <- lme(soi ~ solution + iobs, random = ~1 | iobs/solution,
data
> > stack)
>
> > anova(Lmes.mod)
>
> numDF denDF F-value p-value
>
> (Intercept) 1 2383 2894.8342 <.0001
>
> solution 1 2383 0.0003 0.9870
>
> iobs 1 147 0.0126 0.9109
>
> > summary(glht(Lmes.mod, linfct=mcp(solution="Tukey")))
>
> Error in mcp2matrix(model, linfct = linfct) :
>
> Variable(s) 'solution' of class 'numeric' is/are not
contained as a factor
> in 'model'.
>
This is telling you solution is a numeric variable in stack at this point,
correctly so.
>
>
>
> I don't understand the error since "solution" is clearly a
factor in the
> model.
>
Here's what I don't understand. Why do you believe a procedure that
will produce 17 * 16 / 2 = 136 pairwise comparisons will be scientifically
meaningful? Please tell me you're not using a time variable as an
unordered factor with 17 levels... If you don't understand why that's a
problem, you need to consult with a local statistical expert. Seriously.
HTH,
Dennis
>
> Any suggestions would be welcome.
>
> Bruce
>
>
>
>
>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
[[alternative HTML version deleted]]