I need your help regarding hurdle models, please. Particularly I am interested on 1) whether I have written the command right and 2) how can I interpret the size of the association for the binary exposures if possible.
I'm looking at the association between sedentary behaviour (unit in minutes) and five exposure variables (four of which are binary). Please see more details below:
Independent variables:
1. Speaking on the phone (binary: users and non users)
2. Texting (binary: users and non users)
3. Social networking (binary: users and non users)
4. Internet browsing (binary: users and non users)
5. Composite estimate of time spent in screen behaviour (count variable, I added up the variables 1. Speaking + 2. Texting + 3. Social networking + 4. Internet to create a composite before making them binary)
Dependent Variable:
Time spent in sedentary behaviour (count, lots of 0)
I also want to add five confounders (categorical).
1) The command:
hurdle linear sedentary behaviour i.speaking i.texting i.social networking i.internet i.confounder1 i.confounder2 i.confounder3 i.confounder4 i.confounder5, select(i.speaking
i.texting i.social networking i.internet i.confounder1 i.confounder2 i.confounder3 i.confounder4 i.confounder5
) ll(0)hurdle linear sedentary behaviour icomposite screen behaviour i.confounder1 i.confounder2 i.confounder3 i.confounder4 i.confounder5, select(composite screen behaviour
i.confounder1 i.confounder2 i.confounder3 i.confounder4 i.confounder5
) ll(0)
Do they look right?
2) Can I interpret somehow the size of the association for the binary exposures? I have seen the margin command for the count exposure on how many more or less minutes of sedentary behaviour is associated with the exposure but I don't know whether or how margin command could be used for the binary exposures. Any ideas please?
Thank you very much for your time
Ellie
0 Response to linear hurdle models
Post a Comment