[BioC] 2 sample t test using xps package

[cstrato]

Dear Yonghong

Option "group=" must contain group indices in the order of the 
treenames. This option cannot be used to select treenames. However, you 
can do the following:

Suppose, your rma normalized ExprTreeSet is called "test.rma". Then you 
can create a subset using function "root.expr()". Simply add only those 
treenames which you want to use for further analysis.

You can get the tree names using:
 > treenames <- unlist(treeNames(test.rma))

Regarding your second question, here are two examples:

 > hithreshold=c(10.5,80.0,"percent")
This means that at least 80% of the samples must have a log2 value <= 
10.5 for each probeset to be included in further analysis.

 > lothreshold=c(6.0,0.02,"mean")
This means that only those probesets will be included for further 
analysis, where the trimmed mean "mean(x, trim=0.02)" of the log2 values 
for all samples is >= 6.0

I hope that this could answer your questions.

Best regards
Christian


Wang, Yonghong (NIH/NCI) [C] wrote:
> Dear Christian:
> Thanks for the help to my previous question. So far I have gone through
> all normalization and filtering steps and ready to do the t test. When I
> did the normalization, I use all 30 exon array files, but for the t
> test, I would like to leave some files out, in this case, is there a way
> to code this information in the "unifilter" function "group = " option
> without rerun the script again? Another minor question is about
> "PreFilter" lothreshold and hithreshold options. Even though I have read
> the manual of the package, I still have trouble of understanding the
> detailed definition (or calculation) of these two cutoff. Would you
> please give me a lit bit more explanations.
>
> Thanks again
>
> Best regards
>
> Yonghong 
>
>