Average scales for latent variables
Posted: Fri Mar 04, 2016 8:10 am
I've seen some research study using average scales of a collective measurements as an observed variable for a latent variable. I was planning to use the same method. Is that ok for SEM?
To explain clearly, I currently have 66 measurements. These measurements are expected to be a member of 19 square boxes. So, instead of using the direct measurements to be loaded with Motivation, Opportunity, Ability, and Success, can I use average scales for the square boxes to be observed variables of the construct?
For example:
3 measurements (5-Likert scale) are expected to be a member of the first white square box. The average scale from the 3 measurements is 4.2. Can I use 4.2 as an observed variable for the "Motivation" construct?
To explain clearly, I currently have 66 measurements. These measurements are expected to be a member of 19 square boxes. So, instead of using the direct measurements to be loaded with Motivation, Opportunity, Ability, and Success, can I use average scales for the square boxes to be observed variables of the construct?
For example:
3 measurements (5-Likert scale) are expected to be a member of the first white square box. The average scale from the 3 measurements is 4.2. Can I use 4.2 as an observed variable for the "Motivation" construct?