Location Specific Effects

[Yawo Kokuvi]

Hello, I am estimating a regression model predicting the odds of openness to PREP among foreign born and American born persons at two settings: clinic and community. Foreign born were approached only in the community, while Americans in the clinic.

so the population is stratified into two groups: nativity and location.

i want to be able to account for these differences - specifically whether location or nativity makes a difference in the outcome variable. I can combine all the data and then include dummy variables for each group (foreign vs native) and location (clinic vs community).

Is this the most optimal approach or do we have to do an HLM?

thanks, Yy

[Carlo Lazzaro]

Yawo:
your choice mainly depends on the nesting yes/no design of your study.

[Yawo Kokuvi]

Thanks. Could you please elaborate further regarding “nesting yes/no” design of the study?

thanks - Yawo

[Carlo Lazzaro]

Yawo:
I meant that I would check if persons are nested within hospital and/or communities.

[Yawo Kokuvi]

Ok. I realized that the foreign born are only recruited in communities and the Americans in clinics. So there is some nesting.

apart HLM, is there an overall global test of group differences for logistic regression?

[Carlo Lazzaro]

Yawo:
you can add -i-group- as a predictor in the right-hand side of your regression equation.

[Yawo Kokuvi]

ok ... I was thinking of something similar, so I can add two -i.group: i.location and i.nativity. A significant i.group variable will then tell us that the differences matter. Right?

[Carlo Lazzaro]

Correct.
You can also interact your predictors if it makes (methodological) sense.

[Yawo Kokuvi]

Thanks so much