I was wondering what the difference might be when you create a umx ACEv model and to get the standard estimates (for purposes outside of the likelihood based CIs using umxConfint) you use mxSE on the parameter of interest.

So essentially, doing:

ACE=umxACEv()

umxConfint(ACE, run=TRUE)

^Extract CIs from here (using summary).

Then doing:

mxSE(top.A_std,ACE,run=TRUE)

vs just:

ACE=umxACEv()

mxSE(top.A_std,ACE,run=TRUE)

It appears that by doing it one way or the other I actually get different standard errors. Is there any reason why that might be?

Perhaps additionally important is that I also get different estimates in general from doing it one way or the other? Not sure why that may be, or if it is a product of this issue or related.

EDIT:

I think I made a mistake here! Actually, there doesn't seem to be any difference from what I can tell, just a mix-up. Sorry about that!

One question along the way, though, that I am wondering, is, is there a significant difference between umxConfint and umxCI? It looks like they return essentially (but very marginally different) CIs? And as the CIs are likelihood based, does that mean there are asymmetric SEs, etc. for them, and if so, are those possible to obtain?

Thanks as always!

`mxSE`

gives linear approximations to the standard errors for the free parameters and arbitrary algebraic functions of the free parameters. These standard errors can be used to create Wald-type confidence intervals when the necessary asymptotic normal theory applies. The Wald-type confidence intervals differ in important ways from profile likelihood confidence intervals.`umxCI`

is a variation on the`mxCI`

function for profile likelihood confidence intervals.`umxConfint`

extends the R`confint`

method to OpenMx model objects. It uses profile likelihood confidence intervals.If you adjust the % of the confidence interval from its default of 95% to ~68%, it should be close to mxSE() estimates as long as there are enough data to achieve asymptotic behavior of the SE. If there aren't enough data (and especially near a bound such as -1, 0 or 1) then asymmetric CIs may be found. Comparison of the mxSE and mxCI can therefore be informative.