Van der Sluis in OpenMX?
tvo015
Joined: 01/18/2024
I am trying to run an extended univariate moderation model (van der Sluis et al., 2012) in OpenMX (previously done in MPLUS). I am running into an issue of how to account for both individual and co-twin moderator effects as they differ where the extended univariate moderation model extends the means model such that adjustments are made for both individual and co-twin to account for twin resemblance on the moderator, adjusting for the shared influence between an individual and co-twin's moderator effect and corresponding association with the outcome. Essentially, in the phenotypic means model, we want to make familial adjustments on the moderator to result in different regression coefficient estimates depending on zygosity.
In MPLUS, we did it like this (shown only for MZ-- repeated the process for DZ):
Cognition_1 on Moderator_1*(ModcogMZ);
Cognition_1 on Moderator_2*(Modcog2MZ);
Cognition_2 on Moderator_2*(ModcogMZ);
Cognition_2 on Moderator_1*(Modcog2MZ);
Any insight on how to have this done in OpenMx in a moderation model would be much appreciated! Thanks!
More info
My read of it is
The Cognition_1 variable is predicted by the Moderator_1 variable with a free parameter label of ModcogMZ, and similarly for the rest. If that understanding is correct, then an `mxPath()` example for this would be
mxPath(from=c('Moderator_1', 'Moderator_2', 'Moderator_2', 'Moderator_1'),
to=c('Cognition_1', 'Cognition_1', 'Cognition_2', 'Cognition_2'),
labels=c('ModcogMZ', 'Modcog2MZ', 'ModcogMZ', 'Modcog2MZ'),
free=TRUE)
The idea is similar if you're using `mxMatrix()` too: things with the same label are constrained to be the same.
Log in or register to post comments