# Path Model with (endogenous) Treatment - which model?

I have collected data on the intention to create a new business of my students.

I measured it before the course (t1) and after it (t2). I have data on their absenteeism in class and the time they spent on the online platform. My hypothesis is that following my course will modify their attitude toward new business ventures and thus their intention to create a new business.

The hypothesized model is as follow:

All variables are continuous. What is difficult is that the model includes both panel and structural equation modeling features.

Do you know a class of model that could accommodate this?

In addition, absenteeism and time spent on the online course are probably endogeneous. I would be very interested, if you know of a way to treat this additional issue.

Any pointer would be much appreciated - Thank you very much in advance!

(The question was previously posted without luck on StataList)

• The path model isn't clear (to me). Typically, a box represents a single variable. Attitude(t1, t2) is two variables (isn't it)? You have time spent on online course predicting attitude (t1), but attitude t1 happens before time spent on course so this is predicting backwards in time. – Jeremy Miles Aug 19 at 16:41
• The model is clearer if all your arrows point from left to right, not backwards. – Jeremy Miles Aug 19 at 16:42
• Thank you for the direction of arrows, I will try to do something clearer next time. – Luc Meunier Aug 20 at 15:49
• Attitude (t1,t2) is the same variable measured at two points in time. I displayed it as a single variable as I use long format (in wide format, it would be two variables indeed). – Luc Meunier Aug 20 at 15:51
• Change your data and conceptualization of it to wide format. This will make the causal relations much clearer and then you can simply estimate the model using a structural equation modeling software like gsem in Stata. – Noah Aug 20 at 19:08