I have some questions about the best analysis design for retrospective data. The data are the relationship and marriage biography of a person and a health indicator. I have the data in the long format (person-years) beginning with the birth to age at the interview.
The dependent variable is a dummy variable which shows person-years with bad health. The independent variables are dummy variables which shows the relation or marital status (single, dating relationship, cohabitation or married).

1) Can i use fixed effects regression models to analyize the data?
2) Is there a problem, if the relationship status can change back to single or from cohabitation to married. What does the parameter estimate describe, the effect of the transistion from what relationship status?
3) Should i restrict the sample to investigate only transistion from one status to another (e. g. single to dating partenrship) or can i analyze the different relationship status simultanously

I hope someone can help me.
Best regards
Björn