Moderation analysis using Stata (13): Interaction variables or structural equation modelling (sem)?

[Ourega-Zoe Ejebu]

Dear Statalist users,

I am interested in the relationship between work absenteeism (dependent variable) and work-related stress (independent variable) in various occupational sectors.
I want to explore whether "organisational commitment" alter the strength of the absenteeism/work-related stress relationship, in other words if organisational commitment has a moderator role in this relationship.

The literature I have read thus far suggest that an ordinary regression with interaction effect would be sufficient to test the moderator effect of organisational commitment, such that:

absenteeism= a*work-related stress + b*organisational commitment + c*organisational commitment x work-related stress + error term

Does interaction term indeed constitute a measure of the moderation or is a structural equation model (sem) more suitable?

Thank you

Zoe

[Richard Williams]

As described, I am not sure why you would need a SEM model. sem does allow for multiple group analysis, so that might be handy. However, my first impulse would be to do OLS regression with interaction terms involving occupational sector. Or, possibly, an xtreg or multilevel model, especially if there are a lot of sectors.

[Bruce Weaver]

I agree with Richard (#2).

How is absenteeism measured?

[Ourega-Zoe Ejebu]

Dear Richard and Bruce,
thank you both for your answers.Much appreciated.

There are only 2 sectors (I might expand it to three maximum) so I guess Richard's suggestion of using an interaction term is more appropriate.

Absenteeism is measured as a percentage rate, i.e. managers are asked "what is the percentage rate of days lost due to work-related sickness"?.

Thank you.