percentage of the variance in the data for each LV
as22kk
Posted on 08/11/08 10:47:03
Number of posts: 60
Hello,
can one calculate the percentage of the variance in the data for LVs through task PLS or it can only be calculated through the behavarial pls?
Thanks,
/Alireza
Untitled Post
rmcintosh
Posted on 08/11/08 12:02:09
Number of posts: 394
No, there is no easy way to do this. Since there is no formal error term in PLS, it becomes tricky to derive a measure that is akin to R2 as you would have in regression. Moreover, I actually don't really know that it would be that useful. Consider that in fMRI data, the vast majority of what is measured is some combination of machine noise and physiological noise, and the signal you are interested in is a tiny proportion of this. The usual use of percentage of variance is to get a rough estimate of how important a given effect is in your data, and that is provide by the % cross-block covariance (this is also true form behavior PLS, there is no %variance in any version of PLS that we have). You could envision using %variance in a predictive sense, but that's outside of what our code provides at present.