Hi everyone, after some thorough searching on my own, I think I need a little nudging in a helpful direction:
GOAL:
Derive boostrapped Confidence Intervals for my specified indirect effects while using GSEM for model building.
CURRENT APPROACH:
I am using GSEM for model building (factor variables specified as IV's), and using NLCOM to derive the indirect effects. I would like build a bootstrap program to do this.
So far, the GSEM works without issue.
My NLCOM works without issue.
My attempts thus far to building a functioning bootstrap program for this has been a total failure. The program(s) I have built thus far are more than happy to run GSEM and NLCOM, but my attempts to store the relevant parameter from the NLCOM has been a bust.
My current understanding of what I should be doing is I would like to save the parameter from the NLCOM into a defined scalar, and this defined scalar is then referenced in the boostrap program. However, I fail here, as I always get the error "non r-class program may not set r() ". I presume this means I'm missing something obvious because of my limited Stata knowledge.
I am *not* committed to using NLCOM if there is another way to derive the indirect effect that plays nicely with bootstrap. I *am* committed to sticking with GSEM if at all possible.
Thank you in advance for any tips!
GOAL:
Derive boostrapped Confidence Intervals for my specified indirect effects while using GSEM for model building.
CURRENT APPROACH:
I am using GSEM for model building (factor variables specified as IV's), and using NLCOM to derive the indirect effects. I would like build a bootstrap program to do this.
So far, the GSEM works without issue.
My NLCOM works without issue.
My attempts thus far to building a functioning bootstrap program for this has been a total failure. The program(s) I have built thus far are more than happy to run GSEM and NLCOM, but my attempts to store the relevant parameter from the NLCOM has been a bust.
My current understanding of what I should be doing is I would like to save the parameter from the NLCOM into a defined scalar, and this defined scalar is then referenced in the boostrap program. However, I fail here, as I always get the error "non r-class program may not set r() ". I presume this means I'm missing something obvious because of my limited Stata knowledge.
I am *not* committed to using NLCOM if there is another way to derive the indirect effect that plays nicely with bootstrap. I *am* committed to sticking with GSEM if at all possible.
Thank you in advance for any tips!
