Is it possible to estimate an AE model using the ACE model functions?

Posted on
Picture of user. tbates Joined: 07/31/2009
Forums
Simple question: How do I estimate an AE model using the `umxACE` function?
Replied on Fri, 02/03/2017 - 07:24
Picture of user. tbates Joined: 07/31/2009

Yes, you can estimate an "AE" `umxACE` model!

This is done via `umxModify` and labels - the labels umx applies to every matrix cell/parameter in the model.

in twin models, these labels take the form matrix name + "_" + "r" + row-number + "c" + col-number. So the link from the first latent variable to the second measured variable in the c matrix is labeled `c_r2c1`.

Just run `umxACE` to get your saturated model, then drop the c parameters using `umxModify`

There is an example of this in `?umxACE`

This is a powerful tool. You can provide a single label to update. e.g. `update = "c_r1c1"`, this can be a list of labels.
By default they are set to zero, but you can set them to anything via `values`. By default they are fixed, but you can free variables with `free = T`.

You can even use regular expressions: umxModify(m1, regex = "c_r.c.”) which would drop all free parameters from the c matrix (as long as there are fewer than 10 rows and columns), which is what you want to create an AE model!

then `umxCompare` to compare them

Because I am lazy, you can update the model , rename it, and do the comparison all in 1 line!


m2 = umxModify(m1, regex = "c_r.c.”, name = "AE", comp= TRUE)