Hi there!
I have a dataset that uses parametric regressors at the group level. One of the behavioral measures for one of the subjects looks to be an outlier. The same subject is an outlier with respect to their parameter estimate on the relevant contrast. The FLAME 1+2 result is as expected. Am I reading the Woolrich 2008 paper correctly in that automatic outlier deweighting would do little to offer piece of mind that this single subject is not driving the entire effect? If so, any suggestions as to what to do in this situation (without removing the subject)?
Cheers very much,
Dan
|