[BioC] Re: how many sample controls at least in moderated t statistics?

[Gordon Smyth]

At 01:29 AM 31/05/2005, weinong han wrote:
>thanks much for your help and explanations.
>
>Sure, I have run limma and computed moderated t statistics on 12 samples 
>in one group and
>only 1 in the other. please see attached R script for my dat analysis and 
>QQ_plot.
>QQ plot looks like somewhat abnormal, how to fix it ?any suggestions wil 
>be much appreciated.

The QQ plot of t-statistics is only supposed to look normal when there is 
no differential expression. Deviations fro the line are supposed to 
indicated differential expression.

>How about the results of moderated t statistics on 12 samples? how to get 
>the full list of significanly differentially expressed genes when setting 
>arbitrary p value cut off less than 0.05?

results <- decideTests(fit, adjust="fdr", p=0.05)

To select all up genes:

fit.up <- fit[results[,2] ==1, ]

To select all down genes:

fit.down <- fit[results[,2]== -1, ]

To output to a file:

write.fit(fit.up[,2], file="myresults.txt")

or simply

write(fit.up[,2], file="myresults.txt")

Gordon

>thanks again.
>
>Best Regards