Comparing mortality rate over years

[ram singh]

Dear Statlisters,
I am trying to find : whether there is a "significant change in the mortality rate (death rate)" over the years from 1985-1989 among the new births in a particular geographical region. Please note that the exposure population is different for each year as the mortality rate is among the "new births" each year the region.

I wanted to confirm that a general poisson model such as below would be appropriate to find whether there was a significant change of mortality rates over the years (see the data below)


Model: Year is the explanatory variable

poisson death year, exposure(births) vce(robust) irr



Please consider the following data


----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. To install: ssc install dataex
clear
input int year long births int death
1985 82758 816
1986 83657 707
1987 83990 630
1988 84315 590
1989 84998 374
end

------------------ copy up to and including the previous line ------------------



thanks in advance for your response
regards
RS

[Carlo Lazzaro]

Ram:
I would propose a minor tweak to your code (please see -fvvarlist- for further details on -i.- prefix):

poisson death i.year, exposure(births) vce(robust) irr

As -poisson. often suffers from extradispersion (that, in most cases means overdispersion), you should check if that nuisance affects your results via -estat gof-.

[ram singh]

Thanks a lot Carlo, much appreciated
I will use your suggestion
with regards
Ram