Back to PLS HelpI'm Online
Posted on 01/28/15 13:01:50
Number of posts: 3
Hello,
I have performed three non-rotated task PLS analyses with the default mean-centering option, each contrasting dSPM maps from two different conditions of MEG data. Each analysis has yielded two LVs, of which I believe only the first is real because I selected the default mean-centering option. Oddly, in all three analyses, the probability for the first LV (the difference between the two conditions) is zero. Having never seen a p value of zero in any other kind of statistical analysis, I can't help but worry that this is an error or incorrect result. If anyone could shed some light on the validity of this result I would be most appreciative.
Thank you for your time,
Lauren
Replies:
Posted on 01/29/15 05:59:12
Number of posts: 394
Hello,
I have performed three non-rotated task PLS analyses with the default mean-centering option, each contrasting dSPM maps from two different conditions of MEG data. Each analysis has yielded two LVs, of which I believe only the first is real because I selected the default mean-centering option. Oddly, in all three analyses, the probability for the first LV (the difference between the two conditions) is zero. Having never seen a p value of zero in any other kind of statistical analysis, I can't help but worry that this is an error or incorrect result. If anyone could shed some light on the validity of this result I would be most appreciative.
Thank you for your time,
Lauren
Hi Lauren
Its entirely possible to get p=0 for a permutation test since its an exact estimate of how often random reshuffling of your data produced a singular value equal to or higher than the observed one. Statistical purity would require you to run all permutaitons possible (which can number in the millions) to be certain it is in fact zero, so when you report it, I would write something like P<<0.001
I'm Online
Posted on 01/29/15 09:13:33
Number of posts: 3
Login to reply to this topic.
