[FieldTrip] FT_STATFUN_DEPSAMPLESREGRT Across Subjects For Cluster Based Permutation Test

[James Stieger]

Hello Fieldtrip!

I am hoping to extend the technique outlined in the following page to
behavioral data across time.

http://www.fieldtriptoolbox.org/faq/how_can_i_test_for_correlations_between_neuronal_data_and_quantitative_stimulus_and_behavioural_variables?s[]=regression

I have behavioral data for each subject over a series of experiments and I
want to test whether changes in averaged topographic frequency information
is related to these behavioral changes over time across a population of
subjects. The way I was hoping to code this up is as follows:

subject 1 behavioral data = [0.3 0.6 0.9]
subject 2 behavioral data = [0.2 0.6 0.75]

cfg.statistic = ft_statfun_depsamplesregt;
design(1,:) = [0.3 0.6 0.9 0.2 0.6 0.75]
design(2,:) = [1 1 1 1 2 2 2 2]
cfg.design = design;
cfg.ivar = 1;
cfg.uvar = 2;

I am assuming here that the unit of observation is subjects and I have a
time series of variables for each subject placed in ivar. finally the input
to ft_freqstatistics will be an averaged struct and the powspctrm with have
dimord = subject_channel_frequency_time.

Am I doing this correctly? Will this work? If not, do you have any
suggestions for how to continue?
Additionally some subjects participated in 7 experiments and some
participated in 10. Can I NaN out the missing data, or should I limit this
analysis to the experiments that all subjects participated in?

Thanks for the help!
James
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20180927/14f67f3b/attachment-0001.html>