Back to the main page.

Bug 2866 - create a FAQ on the effect of variable trial numbers on statistical tests

Status NEW
Reported 2015-03-09 14:43:00 +0100
Modified 2015-06-04 17:03:22 +0200
Product: FieldTrip
Component: documentation
Version: unspecified
Hardware: PC
Operating System: Mac OS
Importance: P5 normal
Assigned to: Jim Herring
URL:
Tags:
Depends on:
Blocks:
See also:

Robert Oostenveld - 2015-03-09 14:43:31 +0100

as discussed in the M/EEG lunch meeting


Robert Oostenveld - 2015-03-09 14:48:26 +0100

I copied the material of the meeting to /home/common/meg/meglunch_20150309_power_bias


Jim Herring - 2015-03-12 12:17:25 +0100

I've created an FAQ page and will start filling it with an example of simulated FieldTrip data and tests using the statistics functions.


Jim Herring - 2015-03-12 12:17:56 +0100

FAQ page: http://fieldtrip.fcdonders.nl/faq/how_does_a_difference_in_trial_numbers_per_condition_affect_my_statistical_test I have not created a link on the FAQ page yet. Will do so after I've added some content


Jim Herring - 2015-03-16 15:29:42 +0100

In the example scripts made by Robert and Rene Matlab's statistics toolbox functions signrank and signtest are used which are not supported by out ft_statistics_stats function (which calls functions from the stats toolbox). As it seems easy enough to add support for these functions I will do that as well.


Jim Herring - 2015-03-16 16:01:21 +0100

Sending ft_statistics_stats.m Transmitting file data . Committed revision 10276.


Robert Oostenveld - 2015-05-28 16:44:34 +0200

Hi Jim, I came across http://www.fieldtriptoolbox.org/faq/how_does_a_difference_in_trial_numbers_per_condition_affect_my_statistical_test and noticed it is not yet complete. If it not going to be further extended, please clean it up, remove the warning and finalise it as is. Otherwise please add whatever you had planned and also get it over with.


Jim Herring - 2015-06-04 16:58:37 +0200

Created attachment 713 Profiler


Jim Herring - 2015-06-04 17:03:22 +0200

Hi Robert, I forgot about this after not having looked at this because simulations using FieldTrip structures took so incredibly long. I now ran a simulation with 20 subjects (100 tests using ft_timelockstatistics, 2 conditions, 100 trials per condition, paired t-test). This simulation took 4369s (72 minutes). I ran the profiler and found that more than half! of the time was spent in ft_version called by ft_preampble_provenance (see attached image). This line is causing most of the delays: [status,output] = system('svn info');