Survey data and testing for trend of prevalence ratios

[ggferrante]

Hi everybody.

I am analysing survey data and used a Poisson regression model to obtain adjusted prevalence ratios (PR) of an outcome variable.

The command I used is the following:
xi: svy: poisson trinciati_log i.claeta4 i.sesso_inv i.istruz3cat i.diffeconomiche3cat_inv i.cittadinanza i.ncs_inv i.lavoro2cat_bis, irr

Now, I need to test if there is a statistically significant trend of PR by age, taking into account all the other variables in the model. Can anyone tell me what command I should use to test this hypothesis?

Thanks in advance to those who can help me.

Gianluigi

[Clyde Schechter]

Which of the model variables is age? The names do not suggest any, in my limited understanding of Italian.

[ggferrante]

Claeta4

[Clyde Schechter]

Well, there are several different tests for trend out there. In this case, the simplest way to do it would be:

svy: poisson trinciati_log c.claeta4 i.sesso_inv i.istruz3cat i.diffeconomiche3cat_inv i.cittadinanza i.ncs_inv i.lavoro2cat_bis, irr

Then for your test of trend, you could look at the p-value in the output for the coefficient of claeta4.

Note that I have not used the xi: here. It is not necessary, now that Stata has factor-variable notation. And using -xi:- actually causes problems if you want to follow-up your regression with the -margins- command. While there are still a few commands that don't support factor-variable notation, they are mostly ancient commands whose functions have been incorporated into newer commands that do. There is the occasional odd situation where -xi- still is needed, but they are rare. So it is probably best if you nearly forget you ever knew about -xi-.

[ggferrante]

Thanks a lot, Clyde. It was very helpful. I'll forget about -xi-
Have a nice day.
Gianluigi