SEM modeling ( path model)

[sedki zn]

Dear ll,

I hope you are doing well.

I'm working on my paper and i would like to study the mediating role of M on the relationship between X and Y like it's being showing following:






Is it correct to study it using the sem command??
Is it called structural equations model? If yes, is it SEM path model ?


kind regards

[sedki zn]

Dear Clyde Schechter,

i would appreciate your help

kind regards

[Clyde Schechter]

Assuming that the variables interest and achieve are both continuous, yes, this can be done with SEM. For those who like giving names to particular types of SEM models, yes, this would be called a path model.

By the way, after you run that model with -sem- you can use the -estat teffects- command to get the total, direct, and indirect effects calculated for you.

[sedki zn]

Clyde Schechter
Thank you very much dear professor

[sedki zn]

Dear Clyde Schechter

I have got a reject on my research paper because of the following reason :





Is that make sense?
Kind regards

[Richard Williams]

A peculiar objection, if you ask me. It reminds me of when I was in grad school, and I told another student that they should use SEM, and they replied "I'm writing a substantive dissertation, not a methodological one." Regardless of topic, I think you should choose the method that is most appropriate given your topic and the data you have.

Having said that, the model appears to be just identified, so you could achieve the same results by just running a couple of OLS regressions. estat teffects is nice but you could do the same calculations by hand easily enough.

If missing data is an issue, sem would be nice because of the fiml option. I wish fiml was an option on other commands besides just sem.

A brief overview of the sem command is at

https://www3.nd.edu/~rwilliam/stats2/l95.pdf

For whatever reason, this editor doesn't like SEM, so if SEM isn't critical for your analysis I suppose you should just humor them.

[sedki zn]

A peculiar objection, if you ask me.

For whatever reason, this editor doesn't like SEM, so if SEM isn't critical for your analysis I suppose you should just humor them.

Thank you dear professor for your intervention.

Actually, i have asked the Editor ( who is the AE as well), she said "SEM does not belong in your financial analysis. I know this, as I have been publishing in finance for 34 years."
Expect she is so "humble", i can't understand her answer because i have used a simple path model as a SEM.

Kind regards

[Clyde Schechter]

You know, I try to be moderate in the way I express my statistical opinions, some of which I hold very strongly. But I can't think of any reasonable way to characterize that rejection letter other than with the word "idiotic." The notion that any particular methodology should be banned from any particular discipline is beyond stupid. If the methodology is appropriate to the study design and the actual data collected, and if it can give a reasonable answer to the research question, then there is no reason not to use it.

Richard Williams makes the good point that you can disguise the use of SEM here, getting the same result through other commands. In that regard, SEM is, in this case, just a convenient way to do something that could be done with -regress- and -nlcom- with a few more lines of code. This fact adds even greater weight to my opinion that that editor doesn't know what he/she is talking about.

Added: Crossed with #7. To which I will point out that holding ignorant ideas for 34 years does not make those ideas correct. What an arrogant person! Frankly, I wouldn't waste my time dealing with her any farther and would avoid that journal until a new editor takes her place. (Of course, that's easy for me to say at this point in my career.)