> Dear all,
>
> I fit independent GLMs for a 2x2 factorial problem on the data matrix of
> size 9500 x 12 (genes x arrays) and get 9500 observed t-values using the
> apply() function. Now, I wish to get the permutated p-values. Therefore
> I random sample the class labels and perform the glm fitting to get the
> t-values from which I can get the p-values. This is done using a for()
> loop. Is there a more efficient way to do this. Each loop currently
> takes 5 minutes approximately.
>
Do you use glm.fit()? by calling directly glm.fit() is faster than calling
glm(); glm() itself uses glm.fit(); see ?glm.fit and you can extract the
t-values from the returned object.
I'm not able to tell something about your next questions, sorry.
best,
vito
> More importantly I need to repeat this at least 1000 times which
> requires 3-4 days but the process halts after some time.
>
> To isolate the problem, I rewrote the script with 10 chunks of 100
> loops. The first 2 chunk runs fine and the results are ok but on the
> third (sometimes fourth, fifth or sixth) chunk, I get the following
> error message:
>
> Error in FUN(newX[, i], ...) : subscript out of bounds
> Execution halted
>
> Does R have a "time out" when I use 'R --no-save <
script.file' on the
> UNIX platform ?
>
> I have checked with my system administrator and according to him there
> is no upper limit to process time. I have explicitly removed every
> unneccassary object at the end of each loop to keep the reserve memory.
> I have tried the same on Windows and different chunks sizes and
> different machines. Sometime it runs fine to completion and when it dies
> it does not appear systematic.
>
> Now I am reduced to writing scripts with chunks of 100 loops and then
> collecting the chunks that were successful. I have to repeat the 1000
> loops for many many different experiments and it is getting very
> tedious.
>
> If you have any idea or had similar experience, please let me know.
> Thank you.
>
>
> Regards, Adai.
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help