permutation tests and data
jcrofts
Posted on 02/27/09 07:16:44
Number of posts: 29
j crofts
Hi, is it fair to say that if the permutation tests return sprob = [0, 0.998, 0.998, 0.998, 0.998] that I cannot trust the resulting analysis? Is it also fair to say that the data is not necessarily garbage, simply that I cannot draw any conclusions from the PLS analysis? Or, does it basically tell me my data is essentially noise?
Cheers, Jonathan
Untitled Post
I'm Online
jshen
Posted on 02/27/09 10:49:41
Number of posts: 291
sprob means the probability of permuted values that are greater than the observed values. therefore, higher sprob value (maximum is 1) means less significant. in your case, you have sprob = [0, 0.998, 0.998, 0.998, 0.998], so only the first LV is significant.
Untitled Post
I'm Online
nlobaugh
Posted on 02/27/09 11:00:01
Number of posts: 229
sprob = [0, 0.998, 0.998, 0.998, 0.998] that I cannot trust the resulting analysis?
however, those are quite unusual values... as Jimmy said, it indicates that you only have one significant effect in the data.
If all the cross-block covariance is accounted for by the first LV, I would suggest double checking your datamats to be sure you have not entered in the wrong data.