Dear all,
In a paper we have submitted recently, one suggestion was that we should
“covary out” reaction times to exclude that differences between our 3
conditions are due to reaction time effects (all 3 conditions differ with
respect to RTs) Although i am not
convinced that covarying Rts out is necessarily a good idea, i do think
that i will have to do this due to the reviewers comments.
From previous discussions on this list and from my own limited
understanding, i think there are in principle a number of ways to do this. I
would be very grateful if somebody more knowledgeable than me could inform
me, whether my ideas are correct and if there any other ways to achieve this
(i.e. “covary out” RTs to show that condition differences “can't” be
explained as simple RT effects)
1.)use one event type for all 3 conditions at the first level design with 2
parametric regresors, one coding condition (1 2 3) and the other containing
trial reaction times. This should covary out RTs from condition differences,
right? But how could i test in this design for differences between two
conditions (e.g. 2 vs. 3 and 1 vs. 3 ), which is what i really want?
2.)Model 3 different conditions each with an parametric regressor for trial
reaction times. But this would only control for within condition RTs?
Although i could search for regions correlating with RTs within conditions
for all conditions with a e.g., a conjunction and hopefully find that those
regions do not correspond to those showing condition effect.
3.)use ANCOVAS at the second level group level by creating 3 different
models for all three pairwise comparison between conditions and use the
corresponding mean RT differences as covariate, this should also remove
differences between conditions after differences in RTs have been taken into
account, or am i totally misguided?
4.)Use a model at the second level in which i correlate mean RTs with
activity for all three conditions to identify areas that correlate with RTS
to come to a comparable conclusion as in 2.) But how would i do that? Use a
multiple regression model with mean activity over conditions and 3
regressors for the three condition-specific RTs, or other there any sensible
ways.
I'm also wondering what differences i should expect when using RTs at
covariate at the first single-subject or at the second group level? And are
there any other useful possibilities to “covary out” RTs or “control” for RTs.
Regards, thanks in advance and greetings from Salzburg
Martin Kronbichler
Martin Kronbichler, M.Sc.
---------------------------------------------
Department of Neurology
Center for Neurocognitive Research
Christian-Doppler Clinic, Paracelsus Private Medical University
Ignaz Harrerstr.79, A 5020 Salzburg, Austria
e-mail:[log in to unmask]
Tel.:+43/(0)/662/4483-3966
---------------------------------------------
Department of Psychology
Center for Neurocognitive Research
University of Salzburg
Hellbrunnerstr.34, A 5020 Salzburg, Austria
e-mail:[log in to unmask]
Tel.:+43/(0)/662/8044-5162
Fax:+43/(0)/662/8044-5126
|