[R] Asking for precisions (was : (No subject))

[Emmanuel Charpentier]

Seen on the R-help list :

>Dear friends.
>I have some numbers in percentages all very close together 
>between 60% and 80% as a repeated measures design, 6 measurements per 
>patient (11 ptt) all on to treatments. I think it might be possible to make 
>arc since transformation to obtain more well behaved values but is that in 
>general the best ? At any rate there is a dramatic difference between the 
>verdict obtained using pairwise.wilcox.test as compared to t.test. So far 
>this is not a R specific question - but I guess it will be. 

Troels, some questions :

- Are your "percentage" values really a proportion (e. g. number of "successes"
from N indemendant "trials") or just an easy way to express vartiatoin from a
baseline (quite frequent in biology and medicine) ? The arcsin(sqrt(p))
transformations makes sense only in the former case (where expected residual
variance is 1/(4*N)). In the latter case, use your "percentage" as a
countinuous variable (maybe after variance-stabilizing transformation).

- What are you trying to assess ? Evolution of responses between repetitions ?
Use some chronological series analysis (e. g. autoregressive models). Or
repeated-measures anova model ... Depends of what your experimental schema is.

- (Paired ?) t-test and Wilcoxon test are tests comparing central tendencies
(means) of two samples. You did not state what defines your twoi samples, and
I've trouble understanding what your experimental scheme is ....

I think that a more detailed description of your experiment or observation set
is in order ...

--
Emmanuel Charpentier			Tel :		+33-01 40 27 35 98
Secrétariat scientifique du CEDIT	Fax :		+33-01 40 27 55 65
Direction de la Politique Médicale // Assistance Publique - Hôpitaux de Paris
3, Avenue Victoria // F-75004 Paris /// France
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._