Dear Statalisters,
I work with Stata 9.2.
I would appreciate your advise on how to proceed with the following analysis:
I intend to explore the association of the variables sex, age,
socioeconomic status and migrant status of individuals with their
participation in a domiciliary survey. The sample list consisted of
names and addresses of subjects.
The variable participation has nine possible categories (incorrect
address, non-existent address, moved, died, absent, ill, doesn't speak
the language, refusal, interviewed) that can be hierarchically ordered
in the following fashion:
Firstly, address can be incorrect, non-existent or correct.
If it's correct, the subject may have moved or may have died or may
live at the address.
If the individual actually lives in the address, he/she can be absent
or be at home.
If he/she is at home he/she can be too ill to participate, unable to
answer because he does not know the language of the questionnaire or
able to participate.
Finally, if he/she is able to participate, he/she may do so or refuse.
I've discarded running a -mlogit- with all categories because the
condition of independence of irrelevant alternatives is not met, i.e.
some alternatives can be 'chosen' only if determined alternatives of a
superior level were 'chosen'. I see two possibilities of action:
1) Running multiple -mlogits- for each level, including in each
-mlogit- only individuals who are assigned a category of that level:
a) incorrect address -non-existent address -address correct
b) moved-died-lives in the address (excluding individuals with
incorrect address -non-existent address )
c) absent-at home (excluding individuals with incorrect address
-non-existent address -moved-died)
d) ill-doesn't speak the language-able to participate (excluding
individuals with incorrect address -non-existent
address-moved-died-absent)
e) refusal-interviewed (excluding individuals with incorrect address
-non-existent address-moved-died-absent-ill-doesn´t speak the
language)
The advantage of this possibility is that I can use the strata,
cluster and finite population corrections specifications of the
survey.
2) Running -nlogit-.
I have no experience with nested logits, and after reading the entry
in the Stata manual I'm not sure my analysis is equivalent to the one
proposed in the manual, because in my analysis each alternative in
each level has a different number of levels bellow it: For example,
incorrect address and non-existent address in the first level do not
have any inferior level, but correct address has 4.
Thank you very much for your time and interest.
Angel Rodriguez-Laso
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
