Message/Author 

WenHsu Lin posted on Monday, November 25, 2013  6:04 pm



Hi, I have a modeling question want to ask. Thank you in advance. I have 2 latent variables (L1 and L2). 3 exogenous variables (x1 x2 x3) 1 mediating variable (m1)[nominal with 4 categories] 1 endogenous variable (o1) model (my thinking): I change m1 into three dummy variables m11 m12 m13 MODEL: o1 on x1 x2 x3 L1 L2 m11 m12 m13 L1 L2 on x1 x2 x3 m11 m12 m13 on L1 L2 x1 x2 x3 what is wrong with this? 


The correct way to do this is described in the following paper which is available on the website: Muthén, B. (2011). Applications of causally defined direct and indirect effects in mediation analysis using SEM in Mplus. Your input using WLSMV is an approximate way to do this. 

WenHsu Lin posted on Tuesday, November 26, 2013  9:13 pm



In the paper for nominal mediator.Does the threshold for a particular class indicate the relationship between that particular group and the outcome variable? 

WenHsu Lin posted on Wednesday, November 27, 2013  12:54 am



Sorry, another related but simple question. My first post model, the correlation between m12 and m13 is very high in Mplus output. However, it is only moderate when I use SPSS to estimate this same correlation. Where did I do wrong? 


I don't know what you mean by your first question. Regarding the m12, m13 correlation, note that there are two kinds of correlations for binary variables: Tetrachorics and Pearson. You may be looking at the former in Mplus and the latter in SPSS. 

WenHsu Lin posted on Wednesday, November 27, 2013  9:24 pm



Is the threshold represents the impact of a particular known class on outcome variable? 


Is your outcome categorical and therefore has threshold parameters? Latent class variables don't have threshold parameters. 

Back to top 