Gang Fang's Blog

Note 1: Mental framework of a SEM

A structural equation model consists of two parts: the structural model and the measurement model, shown as below

Can you describe a case study of SEM application?

Case study: this is found in Structural equation modeling in medical research: a primer. Based on the general framework above, a model of positive symptoms, duration of schizophrenia, and dichotic listening is constructed as follow:

Latent variables in this model include positive symptoms and dichotic listening in both conditions. The relationship between a latent variable and its observed variables, in the case of Positive Symptoms, can be described in such language: “Positive symptoms were measured by hallucinations, disorganized thoughts, and unusual thought content”. The numbers on edges are path coefficients, representing the degree of correlation between variables. A important point to note is all coefficients are calculated simultaneously for all endogenous variables rather than sequentially as in regular multiple regression analysis, implying each coefficient is computed while all other variances are taken into account.

Steps in this SEM study:

1. Step 1 (identify the research problem): to understand how duration of schizophrenia and its positive symptoms affect dichotic listening
2. Step 2 (identify the model): based on prior knowledge, the model shown in above was constructed
3. Step 3 (estimate the model): this is to estimate the parameters of the model, in this case, they are the path coefficients. And they were estimated using SEM and numerical results were shown in the graph
4. Step 4 (determine the model’s goodness of fit): this is to determine the model fits the data well. Analysis of this model using SEM indicated it fit the data well.
5. Step 5 (re-specify the model if necessary): not performed

Longitudinal modeling can be used to compare two or more snapshots of the model to understand relationships that aren’t revealed in measurements taken at a single time.