xuetongzhai
commented
1 year ago Hi. Thanks for reporting this. I checked the code and tried to reproduce the bug with a simulation dataset, but I did not see the inconsistence. The two equations you mentioned should be equivalent. If you don't mind emailing me your data and code to xuz19@pitt.edu, that will be helpful for further debug.
Thank you, Xuetong
Describe the bug When running the ROIaverage()-function and save both outputs of the function - the table and the ChannelStats object - the T values shown in the table are not the same as those in the ChannelStats outptut (ChannelStats.tstat).
To Reproduce Steps to reproduce the behavior: I ran a linear mixed effects model (MixedEffects()) to obtain group channelstats from individual glm results. The output of the LME was entered as first argument into ROIaverage().
Expected behavior I assumed that the table-output of ROIaverage should contain the same values as the individual fields of the channelstats output from ROIaverage. While this was the case for e.g. beta-values and degrees of freedom, the t-statistic, including p-values and covariance matrix differs. I can see that in the code, the T-values are defined separately for the table, while they are inferred from covb in the channelstats-object. I suppose that this is where the difference arises from (see screenshots).
Screenshots
