sarah vidal posted on Sunday, January 26, 2014 - 12:42 pm
I conducted CFA on 2 constructs-- construct 1 has 3 latent factors and construct 2 has 4. I wanted to use these factors for follow-up analyses; more specifically, path analysis. I'm not sure about the best way to use the factor scores in path analysis. I know there are pros and cons about different methods and the following methods have been suggested to me:
-average the indicator scores by factors -weighted sum scores -factor scores via CFA
Does using any of these methods significantly affect the results of path analysis? Can I use latent variables in path analysis?
I plan to include other variables in my model-- and these are all observed variables.
I would recommend using neither of your 3 approaches but instead use the latent variables in your path analysis. They can be used together with your other observed variables.
s2014 posted on Sunday, January 26, 2014 - 1:39 pm
Thanks for your quick response. I tried to use the latent variables in my path analysis; however, I got the following message:
WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES. CHECK THE TECH4 OUTPUT FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE Y8DSDELI.
I'm new to MPlus so I'm not sure if it's my model specification or if it's how I coded the outcome variable Y8DSDELI.
The best way to see what is correlated by default is to run the analysis and look at the results.
You fix a factor correlation/covariance to zero by f1 WITH f2@0.
latent1 latent2 on x1 x2 x3 ;
is the same as
latent1 on x1 x2 x3; latent2 on x1 x2 x3;
Tan Bee Li posted on Saturday, July 02, 2016 - 2:41 am
I examined the factor structure of items from a questionnaire. The two models examined were: Model 1: A correlated five-factor model Model 2: A higher-order five factor model whereby a higher-order factor explains the variances of the five first-order factors.
Will the factor scores generated for the five factors be the same for both models? Or will additional variances be extracted from the five factors in model 2?
Is there a reference I can refer to for the formulas?