Open itziknorman opened 4 years ago
Hi! Thanks for the paper, was not aware of it and will have a look at it. The estimator also looks interesting at first glance. If I understand the heteroskedasticity point of Golan et al. is that they argue that different timepoints should have different SE for a predictor. Sounds quite reasonable to me.
In the development branch there is the function uf_se
which currently implements the "naive" approach.
What I don't know is how this interacts with estimation of autocorrelation / whitening which is usually performed in fMRI research.
So far I haven't used single subject SE errors very much.
Thanks for the tip about the uf_se.m function. When working with ECoG signals, each individual recording site is unique and highly informative.. so you want to be able to quantify the effect size at the single electrode level, e.g. to determine which electrodes to include in 2nd order analyses (this is why I need the SE for).
Recently moved the uf_se to the uf_se branch. (just in case someone is looking for the function)
Hi, I wonder if there is a simple way to output along with the betas an estimate of the Standard Error (SE) - that would allow to compute t-stat in each time-point (separately in each predictor). Currently I'm not sure how to compute statistics at the single channel / single time-point level.
There is one study by Golan et al that tackled this issue in a deconvolution analysis applied to intracranial EEG. They refer to a method described in Davidson and Mackinnon (1993):
Any thoughts on how to implement something like this in unfold? Thanks! Itzik