[FieldTrip] Independent samples ANCOVA permutation test question

[psdwyer at ucdavis.edu]

Dear Fieldtrip community,

 

I'm sorry to take your time, but any insights you have regarding this
question would be greatly appreciated:

 

I am trying to run a cluster-based permutation independent samples ANCOVA on
125-channel ERP averages comparing autistic and typically-developing groups.
I'm trying to covary for multiple continuous between-subject variables, like
IQ, age, etc.  I did look around for previous solutions to this issue, but I
couldn't find anything exactly like what I was looking for.

 

I've attached my somewhat slow and inelegant function that tries to call a
MATLAB ANCOVA function to control for 2 covariates in an ANCOVA.  It seems
to yield plausible F-statistics, but the p-values do not seem sensible to me
- they seem grossly inflated or deflated - which makes me wonder what
happens to the covariates when the permutation distribution is being
created.  Based on some comments in the Fieldtrip documentation I thought it
would be possible to supply 2 different covariates on different cvar lines,
e.g.:

cfg.design(2,:)                     = WISC_PRI_IQ;

cfg.design(3,:)                     = chron_age;

cfg.cvar                            = [2,3];

 

However, I'm wondering if that is mistaken?  Is Fieldtrip able to make sense
of two different control variables in the permutations?

 

I also have, in general, no clue what is happening to the cvar in the
permutations and whether using the cvar option even makes sense for these
kinds of between-subjects covariates.  I tried to look at the resampledesign
function, but I'm afraid I was not clever enough to figure out what it does.

 

Thanks so much!

 

Patrick

 

Patrick Dwyer

Ph.D. Student, Developmental Psychology

 <http://riveralab.ucdavis.edu/> Neurocognitive Development Lab

Center for Mind and Brain, UC Davis

 

Personal Blog:  <http://www.autisticscholar.com/>
http://www.autisticscholar.com/

Pronouns: he/him/his

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200928/0aec1d4a/attachment.htm>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pd_statfun_indepsamplesANCOVA_2var.m
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200928/0aec1d4a/attachment.ksh>