SEM - Mediator variable

[Nguyen Manh Hung]

Dear Statalist. I'm struggling with this. I want to use SEM to proof that GIAODICH (trading freq.) mediates QUATUTIN (overconfidence) and DONBAY (margin use). But I can't recreate the result as the picture. (Mentioning the line QUATUTIN->GIAODICH->DONBAY). I want to export the result exactly like that QUATUTIN->GIAODICH->DONBAY. It is necessary to draw all the variables like the picture below or it only needs 3 variables I've mentioned. Please help me.

[Erik Ruzek]

The variables on the left side of the diagram are covariates. To the extent that you can include those or similar variables, you should do so. SEM is a causal analysis and accordingly, you need to adjust for potential confounders to meet the causal assumption (or have some kind of random assignment).

[Nguyen Manh Hung]

Thank you Mr. Ruzek. I found out exactly what I've mentioned on a

blog. And I couldn't find this kind of function on stata. If you have any information please reply me. I would be very grateful

[Erik Ruzek]

After you estimate a SEM in Stata, you can run the postestimation command, estat teffects to get the direct, indirect, and total effects. If you have Stata version 18, you can also use the mediation suite of commands for causal mediation.

[Nguyen Manh Hung]

After you estimate a SEM in Stata, you can run the postestimation command, estat teffects to get the direct, indirect, and total effects. If you have Stata version 18, you can also use the mediation suite of commands for causal mediation.

Thank you for your helping