sdm-help-list > FWE correction stops at 92 % (with summary of all suggested solutions so far)
Showing 1-2 of 2 posts
Jun 29, 2022 08:06 PM | hauser
FWE correction stops at 92 % (with summary of all suggested solutions so far)
Hello all,
perhaps you can help me, I would be very grateful.
The problem:
Although FWE correction is completed using just a few permutations with SDM, version 6.21, I can not run it successfully with 1000 permutations. Other forum-contributers described error messages popping up, like Emily Zhang, but my FWE correction just stops short of completion (92 % to be exact) and stays there for hours without any change (log attached).
There are a few posts about problems with FWE correction, but so far I have not found a solution solving my problem. Still, I will mention all the solutions I could find in this forum, perhaps it helps another person in need and speeds up their search.
Potentially relevant information:
Since Lydia Fortea also asked this Emily Zhang: The study with the smallest sample size in my analysis is 6 participants in control and patient sample each. As asked by Anton Albajes-Eizagirre in another post about FWE correction problems: Overall 19 studies with 638 participants were used to calculate and correct a mean analysis.
Suggested solution 1: Thread-number
I have used a thread number close to the maximum (7; as described in the manual), half of the maximum (4) and smallest possible (1), since decreasing thread number helped Emily Zhang ("I was able to resolve the problem by decreasing the number of threads"), but to no avail.
Suggested solution 2: Missing values in the SDM-table
My supervisor ran the same analysis successfully before (I just want to analyse further), so contrary to the suggestion of Nanfang Pan the textfiles and SDM-table should not be a problem, although I still checked them for missing values.
Suggested solution 3: Changing the version
I tried using different SDM-versions (6.21, 6.22 and 6.11), although it has to be mentioned that with version 6.22 and 6.11 I get different errors. 6.22 gives an error message when trying to perform the next processing steps, asking you to perform mean analysis first, directly after mean analysis, while 6.11 gave me "ERROR: couldn't load imputation file analysis_MyTest/mi/i00046/Dudas_2017_I00046.sdmimage.bin. Error code: -2" after stopping at 90 % mean analysis. I could not find anything wrong with the SDM-table or the coordinate file for the study by Dudas et al. (2017), but this could be due to the chaos which ensued while trying every possible solution with all the different versions on my computer. Also, I tried different computers (laptop and desktop-pc).
Suggested solution 4: Memory allocation
Following the thoughts of Henrik Voß, I put the software and data on the same and different harddrives. Again, to no avail.
perhaps you can help me, I would be very grateful.
The problem:
Although FWE correction is completed using just a few permutations with SDM, version 6.21, I can not run it successfully with 1000 permutations. Other forum-contributers described error messages popping up, like Emily Zhang, but my FWE correction just stops short of completion (92 % to be exact) and stays there for hours without any change (log attached).
There are a few posts about problems with FWE correction, but so far I have not found a solution solving my problem. Still, I will mention all the solutions I could find in this forum, perhaps it helps another person in need and speeds up their search.
Potentially relevant information:
Since Lydia Fortea also asked this Emily Zhang: The study with the smallest sample size in my analysis is 6 participants in control and patient sample each. As asked by Anton Albajes-Eizagirre in another post about FWE correction problems: Overall 19 studies with 638 participants were used to calculate and correct a mean analysis.
Suggested solution 1: Thread-number
I have used a thread number close to the maximum (7; as described in the manual), half of the maximum (4) and smallest possible (1), since decreasing thread number helped Emily Zhang ("I was able to resolve the problem by decreasing the number of threads"), but to no avail.
Suggested solution 2: Missing values in the SDM-table
My supervisor ran the same analysis successfully before (I just want to analyse further), so contrary to the suggestion of Nanfang Pan the textfiles and SDM-table should not be a problem, although I still checked them for missing values.
Suggested solution 3: Changing the version
I tried using different SDM-versions (6.21, 6.22 and 6.11), although it has to be mentioned that with version 6.22 and 6.11 I get different errors. 6.22 gives an error message when trying to perform the next processing steps, asking you to perform mean analysis first, directly after mean analysis, while 6.11 gave me "ERROR: couldn't load imputation file analysis_MyTest/mi/i00046/Dudas_2017_I00046.sdmimage.bin. Error code: -2" after stopping at 90 % mean analysis. I could not find anything wrong with the SDM-table or the coordinate file for the study by Dudas et al. (2017), but this could be due to the chaos which ensued while trying every possible solution with all the different versions on my computer. Also, I tried different computers (laptop and desktop-pc).
Suggested solution 4: Memory allocation
Following the thoughts of Henrik Voß, I put the software and data on the same and different harddrives. Again, to no avail.
Jun 30, 2022 02:06 AM | emilyz
RE: FWE correction stops at 92 % (with summary of all suggested solutions so far)
Hi! This is Emily Zhang! Even though decreasing the number of
threads seemed to work at first, I ultimately had to buy a faster
computer (with 8 cores and 16 threads) in order to get the FWE
correction to work. Otherwise, there were always error messages.