Dear list members,
I extensively searched in the previous threads of this mailing list an example
easy to understand for me,
and able to fit my problem, but I didn´t succed to find a solution for which I
feel certain.
At the moment I am stuck with this solution for my unbalanced design, and I
don´t know if it is correct or not:
library(nlme)
scrd.lme <- lme(response~stimulus*condition,random=~1|subject,data=scrd)
Now at this point is it correct to simply run the command anova(scrd.lme) ? Or
should I do something different using aov?
As I told, for a balanced case I would use the command
> aov1 = aov(response ~ stimulus*condition +
> Error(subject/(stimulus*condition)),
>data=scrd)
Now in the R prompt I get this output, which is very different from the one
listed of aov for a balanced case:
> anova(scrd.lme)
numDF denDF F-value p-value
(Intercept) 1 182 178.56833 <.0001
stimulus 6 182 1.57851 0.1557
condition 1 182 39.68822 <.0001
stimulus:condition 6 182 0.67992 0.6660
Now, (if the previous point is correct) I am burning with curiosity to solve
another problem.
I found that for condition there are significatively differences.
For condition I have only 2 levels so there is no need to do a post-hoc
analysis.
But if I had 4 levels, I would need.
Now, which is for the ANOVA with repeated measures with UNBALANCED design the
right
approach for a post hoc test?
Is there anyone who can provide a R code example to solve this problem so I can
better understand?
I know, I should read some books to understand better the subject. I am doing
my
best.
Thanks for any suggestion
________________________________
From: Ben Bolker <bbol...@gmail.com>
To: r-h...@stat.math.ethz.ch
Sent: Sat, January 8, 2011 9:39:20 PM
Subject: Re: [R] Anova with repeated measures for unbalanced design
Frodo Jedi <frodo.jedi <at> yahoo.com> writes:
>
> Dear all,
> I need an help because I am really not able to find over
> internet a good example
> in R to analyze an unbalanced table with Anova with
> repeated measures.
> For unbalanced table I mean that the questions are
> not answered all by the same
> number of subjects.
>
> For a balanced case I would use the command
>
> aov1 = aov(response ~ stimulus*condition +
> Error(subject/(stimulus*condition)),
> data=scrd)
I recommend that you find a copy of Pinheiro and Bates 2000 ...
>
> Does the same code still work for unbalanced design?
No, it doesn't.
> Can anyone provide a R example of code in order to get the same analysis?
Something like
library(nlme)
lme1 <- lme(response~stimulus*condition,random=~1|subject,data=scrd)
or possibly
lme1 <- lme(response~stimulus*condition,random=~stimulus*condition|subject,
data=scrd)
if your experimental design would allow you to estimate stimulus and condition
effects for each subject.
Further questions along these lines should probably go to
the r-sig-mixed-model mailing list.
______________________________________________
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]]
______________________________________________
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.
