 |
 |
Hi David, I don't know the reason but I think I would have a look into the ranges in which the images lie in, and have a look into the commands invoked by easythresh (it's a script) to see in which step things become odd with this particular input image. This isn't a helpful answer, I know. If you can't find the problem and would like to send the image (and have a lot of patience to wait) I could have a look in a few days. All the best, Anderson On 8 December 2016 at 10:19, David Hall <[log in to unmask]> wrote: > hey Anderson, > > thanks a lot! I have a follow-up question: when running easythresh with > z>2.3 p<0.05 some of the images turn out to be completely yellow (see > attached picture). This is mostly the case for f-tests or two-sample > t-tests, as one-sample t-tests (average group activation) turn out fine. I > use a gray matter mask. Do you have an idea what this means/what the > problem might be? > > Thanks a lot! > > On Dec 6, 2016 11:24, "Anderson M. Winkler" <[log in to unmask]> > wrote: > >> Hi David, >> >> For various reasons, higher thresholds are better (e.g., high thresholds >> are assumed by the random field theory and give higher spatial specificity >> even for non-parametric tests). >> >> The 2.3 has become common for being the z-score that gives a p=0.01. >> >> Using r=0.95 (or any other value) as a general rule wouldn't be helpful >> as the r depends on sample size. And specifically 1.8 is too low (the RFT >> does not work at that level, for instance). >> >> Instead of lowering, considering increasing to, e.g., 3.1. >> >> All the best, >> >> Anderson >> >> >> On 5 December 2016 at 12:50, David Hall <[log in to unmask]> wrote: >> >>> Dear FSL experts, >>> >>> >>> I have a problem understanding why it is the common practice to use >>> z>2.3 in multiple comparison correction (e.g. easythresh). 2.3 corresponds >>> to the r value 0.98 and I would expect 0.95 to be the most common value to >>> use (i.e. z score of 1.8 ). Is it okay to use 1.8? >>> >>> And another question: when looking at the maps thresholded by easythresh >>> in FSLView, why is it necessary to threshold FSLView again (so that it's >>> E.g. 2.3-3) if theoretically the maps are already thresholded? >>> >>> thanks a lot! >>> >> >>