jpiironen / dimreduce Goto Github PK

Hi,

thank you for this very handy and useful package! I have been using it for extracting features from sequencing data for classification purposes and most of the time the package has worked well. But occasionally the ispca function fails and gives the following error message

"[1] "Computing the supervised principal components.."
Error in FUN(X[[i]], ...) :
Something went wrong: the active set became empty after screening.
Calls: function_that_calls_ISPCA -> ispca -> lapply -> FUN -> spcs -> lapply -> FUN
Execution halted"

function_that_calls_ISPCA is a function in my pipeline which calls ispca function from the dimreduce package. I believe that the source of this error are lines 78-81 from auxiliary.R, where variables that have score above a certain threshold are picked. Based on my debugging, there is probably a problem with floating point precision in this part of the code from line 80

cand[scores[cand] >= th]

which makes the comparison fail.

To be able to run the ispca function to my data, I fixed the problem locally by replacing the part above with

cand[which(round(scores[cand] - th,10) >= 0)]

With this change, I am able to run the ispca function to all my data without problems. It might not be the best way to fix it, but it seems to work. Unfortunately I cannot share the data set for reproducing the error as it is not a public data set.

Best regards,
Viivi Halla-aho

Hi,

First of all thanks for this really handy package! I'm sure I'm going to rely on it in my future projects.

I have a question regarding the P-values of featurescore.test. Isn't it necessary to adjust it for multiple testing?

Best,
Mohamed