model: y1 on y2 x1 x2 ; y2 on x3 ; x1 with x2; x2 with x3; x2 with x3; x1; x2;
1. As you can see, x1 x2 (exogenous) are only related y1 (endogenous). I have found the the loading for x1 x2 are not significant. Do you think it is ok to remove these two variables but to keep the correlations (x1 with x2;), (x2 with x3;), and (x2 with x3;)?
2. If I want to remove the two variables, should I remove the related correlations as well?