Mass Univariate ERP Toolbox: Revision History
The Mass Univariate Toolbox is currently in a "beta" state. Although it has been tested on data from multiple labs, we expect to discover some bugs once more labs begin using it. In addition, we hope to add further functionality to the code in the near future. This page contains a chronological list of changes to the code. To be automatically notified of toolbox updates, go to the toolbox download page, register for a free Mathworks account (if you have not done so already), login, and click "Watch this File":
March 5, 2012: sig_raster.m now can use a temperature scale to represent graded values of significance (in microvolts or t-scores). save_matmk.m supports paths with spaces when using linux thanks to Gábor Háden.
Dec. 12, 2011: find_clusters.m made recursiveless by Amy Guthormsen of Los Alamos National Laboratory's Applied Modern Physics Group. This makes it possible to use cluster-based permutation tests on a much larger number of comparisons (e.g., 128 channels by 200 time points).
Sept. 5, 2011: bin_info2EEG.m fixed. New 'bin*' elements of EEG.events had the wrong bin #. Also, the code wouldn't work if EEG.epoch(#).eventtype was numeric. These problems should not have affected any results from the Mass Univariate Toolbox since it ignores the EEG.events field for the most part. However, it could potentially affect EEGLAB code that uses the EEG.events field. Thanks to Haibo "Seapsy" Zhou for finding these problems and their solutions.
June 27, 2011: Cluster-based permutation tests added to toolbox. Function for GFWER control added.
May 24, 2011: save_erp.m and save_matmk.m were not fully compatible with Windows. Now they should be. Thanks to Elisa Filevich for finding this bug (and the solution). bin_opGRP.m had a bug in which it would not work properly if Group A and Group B did not correspond to the first and second group in the GRP variable (respectively). This has been fixed.
Apr. 22, 2011: bin_info2EEG.m can now work on EEG variables without an EEG.event(x).duration field. tfdrGND and tfdrGRP store FDR corrected p-values in GND and GRP variable for BH and BY FDR control algorithms. tmaxGND and tmaxGRP report min AND max significant p-values to command line.
Apr. 11, 2011: save_matmk.m was added to the toolbox download. It was accidentally omitted from the previous zip file.
Apr. 4, 2011: The toolbox was made compatible with Windows. Specifically, the following functions were slightly changed:
Mar. 23, 2011: decimateGND.m now works on non-Kutaslab data and 'save_GND' option was added to it.
Mar. 10, 2011: gui_erp.m 'show_wind' option was ineffectual but now it works. bin_info2EEG.m was improved and can now handle continuous set files. remove_artifact_ics.m now works if ICA hasn’t been run on the data. gui_erp.m now checks to make sure a bin has data in it before trying to visualize it. sig_raster.m properly deals with channels with a radius coordinate of 0. Some functions were modified to be able to handle power spectrum analysis. Those changes should be invisible to non-Kutaslab users. Thanks to Udo Will for identifying and helping to fix some of these bugs.
Dec. 10, 2010: mxt_perm1.m had a minor problem in that it computed t-scores for the observed data and for permutations of the observed data using different, mathematically equivalent procedures The different procedures could produce slightly different t-scores for the same data due to rounding errors. These rounding errors could produce inaccurate within-subject t-test p-values for very small sample sizes (e.g., 3 participants). That problem has been fixed. Minor changes were made to a handful of other functions as well.
Dec. 4, 2010: sig_wave.m and plot_wave.m were added to the toolbox to help data visualization. A few minor modifications to other functions were made as well.
Nov. 15, 2010: Function erplab2GND.m was added so that this code can be used with the ERPLAB MATLAB toolbox. A few minor modifications to other functions were made as well.
Jan. 24, 2015: Minor bug in tfdrGRP.m fixed. The function was reporting non-adjusted p-values in the command window as adjusted. Note that GRP.t_tests and t-test plots (e.g., those produced by sig_topo.m) were accurate. A minor bug in topoplotMK.m was also fixed. The function could incorrectly display which electrodes exhibited significant differences when some electrodes did not have coordinates and were not displayed.