2

Good afternoon. I am working on a study to determine if there is a difference in caregiver empowerment for those who have worked with a Family Support Partner and those that have not. I am using the Family Empowerment Scale as my measure. To score the measure, I add up the answers that fall into three subsets: Family, Service System and Community/Political. I have been advised this will make my study have three dependent variables as a result. I have 5 independent variables (ethnicity, age, gender, time as a caregiver, time you have worked with a Family Support Partner). The Family Empowerment Scale uses a likert system. I thought about using Ordinal Logistic Regression, but I don't think I can use this test with three dependent variables. Any suggestions?

Thank you

Kelli Conley Taylor
  • 21

2 Answers2

2

As an exploratory, I'd try 3 ordinary least squares regressions, with each of the "independent" variables plus your program participation intervention affecting each if the three dependent variables. This strategy will have the benefit of easily interpretable coeficient outputs for your intervention and demographic variables - ie: "an extra year of age is associated with .02 points increase in the metric of interest".

Armed with this knowledge, you can move onto more complex methods.

Eduard Gelman
  • 161
1

There is an R package, mvord, which can perform multivariate ordinal regression. That might be able to handle your situation?

NatWH
  • 539