Confidence intervals for mortality rates

[Joe Ward]

Dear Statalist

I would like to display confidence intervals around a mortality rate estimate.

I am calculating mortality rate by: number of deaths in 1 year / population in that year per 100000

Some of the number of deaths are not whole numbers (because I am using lagged means over a three year period)

In one example I have 31.965015 deaths in population of 3868012

So mortality rate per 100000 = 31.965015 / (3868012 /100000) = 0.82639389

I think i need to use

ci number_of_deaths , poisson exposure(population_per_100000)

This gives an upper and lower limit, but slightly different estimate of 0.8272984 (using the example above). Here is the output:


-- Poisson Exact --
Variable | Exposure Mean Std. Err. [95% Conf. Interval]
-------------+---------------------------------------------------------------
mean_deat~19 | 38.68012 .8272984 .1462471 .5658715 1.167898

Does anyone know why this is the case?

Many Thanks

Joe

[Joe Ward]

Hello all

For anyone who is interested - the slight difference was due to the use (or lack of) of double precision. I reran the data using cii means population deaths, poisson, and then recast the result as double - which worked

Best Wishes

Joe