I recently had a reviewer comment that non-significant paths in my path model should not be deleted from the model and instead they should be set to zero.
I thought (perhaps erroneously) that regression coefficients are automatically set to zero in Mplus by default unless they are specified otherwise in the model.
I've run the path model both ways (1. only specifying significant paths, 2. specifying significant paths and setting all non-significant paths to zero). The path coefficients and model fit indices only change very slightly between the two models.
My questions are:
1) Which method is typically preferred and which papers should I refer to?
2) How does Mplus handle paths that are unspecified vs. setting them to zero?