I want to find out if there is effect of marital transition (from married to widowed) on self assessed health with 3 wave panel data (unbalanced panel). My question is: What is the correct way to find this out?

Would it be reshaping panel from long to wide format and regressing change in marital status on health with controlling for health in T1? This is the way I have encountered in research articles in this area.

So my model would be something like: healthT3 = transitionT2 + transitionT3 + sex + age + healthT1/T2 + error term

Or should I use fixed effects panel regression? AFAIK fixed effects only counts for changes in variables (120 changes from T1 to T3 married to widowed). So, can FE estimates be interpreted in the sense of change in marital status?

Or is there any other way how to find this out?

  • $\begingroup$ what is health T1 ? $\endgroup$ – Subhash C. Davar Aug 1 '15 at 12:13
  • $\begingroup$ I am not sure I completely get your question. But health T1 is ordinal (5 point scale) variable surveyed in 2006 (T3 is 2013). $\endgroup$ – lvm3n Aug 2 '15 at 19:07

From what I understand, I think using as a starting point using a Fixed effects regression of the following form might be helpful: $H_{it} = \alpha_{i} + \delta_{t} + \beta Maritalstat_{it} + \varepsilon_{it}$ which is the simplest regression that can be run. You can perhaps even interact the $Maritalstat_{it}$ with age variable,to see whether being a widow at an older age is much more harmful (or not) than when one is younger. Here your marital status variable is simply a categorical variable that can be coded for different categories, single, married and widowed. And then you can interpret it as a change in marital status. Hope this helps!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.