| Re: To perform anova for count data in stata with survey weight [message #15755 is a reply to message #15544] |
Mon, 10 September 2018 20:39  |
Bridgette-DHS
Messages: 3230
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Specialist, Tom Pullum:
Sorry for the delay in this response. You may already have solved the problem.
I don't see any obvious reason why you are getting an error message. Can you send a log file?
Personally, I would use Poisson regression and a model such as "poisson no_of_antenatal_visits i.education_level". With anova, you are treating the count as an interval-level outcome. The tests in anova assume that the errors are normally distributed and homoscedastic. Both of those assumptions are implausible for a count. Poisson (and negative binomial) models are more appropriate for an outcome that is a count.
I also recommend that other variables be included, such as the age and/or parity of the woman.
|
|
|
|
Members
Search
Help
Register
Login
Home
