Dear list, im facing an issue of statistical data analysis that I consider myself unable to resolve in R so i hope to get some valuable insights from you. i run an ANOVA with four factors; factor4 is an between factor (two different groups measured), the others are withins (tested across /all/ subjects). accordingly, my model looks as follows: fm1 <-aov(dataSubjects~(factor1*factor2*factor3*factor4)+Error(subject/(factor1* factor2*factor3))+factor4,data=mydata) print(summary(fm1)) so far so good. i now would like to perform multiple comparisons, but separately for each level of factor4 (two). this is how I do it: i sort out (using subset) the levels im interested in comparing, say level 1 of factor 1 Type.lvl1 <- subset(mydata, factor1 == "lvl1") and then compute an uncorrected pairwise t-test between the two levels of factor 4: Type.lvl1.P<-pairwise.t.test(Type.lvl1$dataSubjects,Type.lvl1$factor4,p.adj = "none") i do this for every comparison im interested in and apply any method for multiple comparison correction afterwards on the entire set of p values obtained. is this valid (in statistical terms)? if not, could anyone refer to a better procedure i should use instead? Thanks in advance. Lisa WEB.DE DSL Doppel-Flat ab 19,99 ?/mtl.! Jetzt mit gratis Handy-Flat! [1]http://produkte.web.de/go/DSL_Doppel_Flatrate/2 References 1. http://produkte.web.de/go/DSL_Doppel_Flatrate/2
Lisa,
Please look at some of the demos in the HH package.
These are built on the capabilities of the glht function in the multcomp
package.
## install.packages("HH") ## if necessary
library(HH)
demo("MMC.WoodEnergy-aov", package="HH") ## first
demo("MMC.WoodEnergy", package="HH") ## second
Rich
On Wed, Apr 27, 2011 at 5:02 AM, <elgoran@web.de> wrote:
>
> Dear list,
> im facing an issue of statistical data analysis that I consider myself
> unable to resolve in R so i hope to get some valuable insights from you.
> i
> run an ANOVA with four factors; factor4 is an between factor (two
> different
> groups measured), the others are withins (tested across /all/ subjects).
> accordingly, my model looks as follows:
> fm1
>
>
<-aov(dataSubjects~(factor1*factor2*factor3*factor4)+Error(subject/(factor1*
> factor2*factor3))+factor4,data=mydata)
> print(summary(fm1))
> so far so good. i now would like to perform multiple comparisons, but
> separately for each level of factor4 (two). this is how I do it: i sort
> out
> (using subset) the levels im interested in comparing, say level 1 of
> factor
> 1
> Type.lvl1 <- subset(mydata, factor1 == "lvl1")
> and then compute an uncorrected pairwise t-test between the two levels of
> factor 4:
>
>
Type.lvl1.P<-pairwise.t.test(Type.lvl1$dataSubjects,Type.lvl1$factor4,p.adj
> = "none")
> i do this for every comparison im interested in and apply any method for
> multiple comparison correction afterwards on the entire set of p values
> obtained. is this valid (in statistical terms)? if not, could anyone
> refer
> to a better procedure i should use instead?
> Thanks in advance.
> Lisa
>
> WEB.DE <http://web.de/> DSL Doppel-Flat ab 19,99 €/mtl.! Jetzt mit
> gratis Handy-Flat! [1]http://produkte.web.de/go/DSL_Doppel_Flatrate/2
>
> References
>
> 1. http://produkte.web.de/go/DSL_Doppel_Flatrate/2
> ______________________________________________
> 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<http://www.r-project.org/posting-guide.html>
> and provide commented, minimal, self-contained, reproducible code.
>
[[alternative HTML version deleted]]
Rich,
thanks a lot, i will definitly check it out. however, since the analysis
mentioned above is already implemented could you or anyone tell me whether
it contains any statistical flaws?
best
Lisa
______________________________________________________________
Von: "Richard M. Heiberger" <rmh at temple.edu>
Gesendet: 27.04.2011 16:58:43
An: elgoran at web.de
Betreff: Re: [R] multiple comparisons on a between factor
Lisa,
Please look at some of the demos in the HH package.
These are built on the capabilities of the glht function in the multcomp
package.
## install.packages("HH") ## if necessary
library(HH)
demo("MMC.WoodEnergy-aov", package="HH") ## first
demo("MMC.WoodEnergy", package="HH") ## second
Rich
On Wed, Apr 27, 2011 at 5:02 AM, <[1]elgoran at web.de> wrote:
Dear list,
im facing an issue of statistical data analysis that I consider myself
unable to resolve in R so i hope to get some valuable insights from you.
i
run an ANOVA with four factors; factor4 is an between factor (two
different
groups measured), the others are withins (tested across /all/ subjects).
accordingly, my model looks as follows:
fm1
<-aov(dataSubjects~(factor1*factor2*factor3*factor4)+Error(subject/(factor
1*
factor2*factor3))+factor4,data=mydata)
print(summary(fm1))
so far so good. i now would like to perform multiple comparisons, but
separately for each level of factor4 (two). this is how I do it: i sort
out
(using subset) the levels im interested in comparing, say level 1 of
factor
1
Type.lvl1 <- subset(mydata, factor1 == "lvl1")
and then compute an uncorrected pairwise t-test between the two levels
of
factor 4:
Type.lvl1.P<-pairwise.t.test(Type.lvl1$dataSubjects,Type.lvl1$factor4,p.ad
j
= "none")
i do this for every comparison im interested in and apply any method for
multiple comparison correction afterwards on the entire set of p values
obtained. is this valid (in statistical terms)? if not, could anyone
refer
to a better procedure i should use instead?
Thanks in advance.
Lisa
[2]WEB.DE DSL Doppel-Flat ab 19,99 ?/mtl.! Jetzt mit
gratis Handy-Flat! [1][3]http://produkte.web.de/go/DSL_Doppel_Flatrate/2
References
1. [4]http://produkte.web.de/go/DSL_Doppel_Flatrate/2
______________________________________________
[5]R-help at r-project.org mailing list
[6]https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
[7]http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Empfehlen Sie WEB.DE DSL Ihren Freunden und Bekannten und wir
belohnen Sie mit bis zu 50,- Euro! [8]https://freundschaftswerbung.web.de
References
1. mailto:elgoran at web.de
2. http://web.de/
3. http://produkte.web.de/go/DSL_Doppel_Flatrate/2
4. http://produkte.web.de/go/DSL_Doppel_Flatrate/2
5. mailto:R-help at r-project.org
6. https://stat.ethz.ch/mailman/listinfo/r-help
7. http://www.r-project.org/posting-guide.html
8. https://freundschaftswerbung.web.de/