[julien.sylvestre@wotan.ens.fr: Re: [BioC] pb using affy // Anova with R]

[Laurent Gautier]

Forget to CC to the list (and could be relevant to others)



L.



03/05/2002 14:55:35, Laurent Gautier <laurent@genome.cbs.dtu.dk> a ?crit:

>On Fri, May 03, 2002 at 03:22:16PM +0200, julien.sylvestre@wotan.ens.fr wrote:
>> * I've been using the affy package for a few days, and found it very well made, and really useful. By the way, thanks to all of yours for  building a well documented and coherent software collection, which 
>> remains somewhat rare in the microarray domain, everyone tending to write it's own little program...
>> 
>> * Yesterday, after switching to the 1.0 version (on Win2K / R1.5.0) , I experienced the following problem  while fitting  the  Li & Wong Model :
>> //  liw2 = express(pl1, normalize = F, summary.stat = li.wong)
>> // Background correcting
>> // Preparing Data
>> // Computing expression. This may take a while.
>> // Error in data.matrix[, !phi.outliers] %*% phi[!phi.outliers] : 
>> //  non-conformable arguments
>
>
>
>In does not seem the code used/called by 'express' was modified recently. Could you specify with
>which version it was working ?
>
>What gives 'traceback()' ? (type that right after you had the error) 
>
>> 
>> What does this error mean ? 
>
>that we have an... <embarassed cough>... undocumented feature... ;-)
>
>
>> I had no problem when applying the default summary.stat with affy 1.0 and the same plobe level object. 
>> I had no problem either with the former version, the same plob and Li.Wong .
>> 
>
>Do you get the same when going through the Cel.container mecanism ?
>
>
>
>
>
>> Thank you in advance.
>> 
>> JS,
>> 
>> * PS. One minor thing maybe somebody can help me with :
>> I have 12 yeast chips to analyse and the experimental design is quite well suited for ANOVA (1 factor A with 2 levels, 1 factor B with 3 levels, 2 reps C for each).
>> I tried 'lm' and 'aov' on R but encountered serious memory problems despite setting memory.limit(size = 4000) and having 1GoRam plus a lot of free hard drive.
>> I tried 'anovan' from Matlab Stat Toolbox, which of course resulted in a complete failure.
>> In fact I'm not necesserally interested in fitting a whole model, with thousands of gene effects, but just in calculating the average A, B , C , A*B , A*C , B*C   effects. I know it can be done with proc anova on 
>> SAS but if it were possible, would prefer to do the whole thing under R. Does anyone know any package or function I could use ?
>> 
>
>?!
>Could you give us more details about the way you proceed ?
>(note: the code of the function 'Anova' in the package 'genefilter' could be helpful)
>
>
>
>
>Cheers,
>
>
>
>Laurent
>
>
>
>
>--------------------------------------------------------------
>Laurent Gautier			CBS, Building 208, DTU
>PhD. Student			D-2800 Lyngby,Denmark	
>tel: +45 45 25 24 85		http://www.cbs.dtu.dk/laurent
>
**************

* Je n'ai pas essay? le 'Cel.container mecanism' ; ce qui me para?t ?trange c'est que cel? fonctionnait avant (version 0.8) et que ?a fonctionne avec d'autres m?thodes. 
Pour information, j'obtiens ?a en tapant 'traceback()':
6: fit.li.wong(t(data.matrix), remove.outliers, normal.array.quantile, 
       normal.resid.quantile, large.threshold, large.variation, 
       outlier.fraction, delta, maxit, outer.maxit, verbose)
5: FUN(X[[3]], ...)
4: lapply(as.list(X), FUN, ...)
3: sapply(data.lst, summary.stat, ...)
2: t(sapply(data.lst, summary.stat, ...))
1: express(pl1, normalize = F, summary.stat = li.wong)

* Je vais regarder genefilter, en fait je pensais que c'?tait tr?s orient? puces cDNA . Effectivement, l'id?al serait davoir une ANOVA g?ne par g?ne.