Which are the dependent and independent variables in a Structural Equation Modeling (SEM)?

#1
Hi, everyone, I am reading some literature now, and I need to do a research with Structural Equation Modeling(SEM); however, unlike normal research model, which only involved with one dependent variable (C) and one independent variable(A), or plus one mediation variable(B), like A>B>C. there are always lots of variable in a SEM. I am wondering how do I know which variable in the SME model is independent variable, and which are the dependent variables in a SEM model below.
 
Last edited:
#4
a) A normal model does NOT necessarily have only one independent variable, it can have many.
b) mediating variable is not present in all statistical models. It is rarely used, depending on the problem.
c) in SEM, you define beforehand the hypotheses you are going to test. Accordingly you identify the variables BEFORE you decide on the test. the order of this process is as follows:

1) Studying the problem
2) identifying the independent and dependent variables
3) frame the hypothesis
4) decide the statistical test to apply (can be SEM/ regression/ chi square, etc.)
5) apply the test.

Step 2 and 3 can be interchanged.
Most importantly, SEM needs identification of LATENT and MEASURED variables. Ive talked more about it here: https://www.projectguru.in/understanding-the-application-of-structural-equation-modelling-sem/

Good luck!
 
#5
a) A normal model does NOT necessarily have only one independent variable, it can have many.
b) mediating variable is not present in all statistical models. It is rarely used, depending on the problem.
c) in SEM, you define beforehand the hypotheses you are going to test. Accordingly you identify the variables BEFORE you decide on the test. the order of this process is as follows:

1) Studying the problem
2) identifying the independent and dependent variables
3) frame the hypothesis
4) decide the statistical test to apply (can be SEM/ regression/ chi square, etc.)
5) apply the test.

Step 2 and 3 can be interchanged.
Most importantly, SEM needs identification of LATENT and MEASURED variables. Ive talked more about it here: https://www.projectguru.in/understanding-the-application-of-structural-equation-modelling-sem/

Good luck!
many thanks to your reply, Hello, Everyone! I am wondering if there is only one way to test a model? As you can see the model below, for example, instead of using one hypothesis (H1) to test the linear effect from variable 1 to variable 4 through structural equation modeling (SEM) , could I make three hypotheses to test the same relationship by regression analysis, first , I test the relationship between variable 1 and variable 2 (H1), and then the relationship between variable 2 and variable 3( H2), and finally the relationship between variable 3 and variable 4 (H3)? and due to equal exchange; therefore, I can say that there is positive effective between H1 and H4.talking about which test I am suppose to use, may I ask you one more question? 1633717980889.png