Structural Equation Models

CARLOTA

New Member
Hello!

I am new in this forum.

My question is because in my PhD I am performing SEM with second order factors. I am using SAS system for analysis.

My model consists in 2 second order factors that each explain 7 first order factors, which those explain 20 manifest variables.
We have fixed one factor loading (one path from one of the first order factors towards one manifest variable) to be at 1. And now my question is about the second order factors. Should I estimate their variances (as I do for errors) or should I fix the variance of those second order factors at 1?

Carlota

CARLOTA

New Member
A professor in statistics has given me the answer. I post it here, in case someone is interested:

"I typically will fix one loading from the higher-order factor to the first order factors to one as a way of scaling the higher order factor to be in a similar metric to that of the first order factor that has its coefficient set to one. I also set one loading per first order factor to one as a way of scaling the first order factors. However, it also is possible to set the second order factor's variance to 1 as an alternative way to scale it."

CARLOTA

New Member
Fix the variance to 1

A professor in statistics has given me the answer. I post it here, in case someone is interested:

"I typically will fix one loading from the higher-order factor to the first order factors to one as a way of scaling the higher order factor to be in a similar metric to that of the first order factor that has its coefficient set to one. I also set one loading per first order factor to one as a way of scaling the first order factors. However, it also is possible to set the second order factor's variance to 1 as an alternative way to scale it."