Writing Estimate Statements and Interpreting Poisson Output

EpiNovice — Wed, 13 Nov 2019 16:13:06 GMT

Hello everyone,

I'm trying to understand estimate statements and I'm really struggling, so I hope you can assist!

My variables are as follows:

Outcome: incidence of events per person-year

Disease (yes=1, no=0)

Sex (male= 1, female=0)

Study site (L= 1, M=3)

Season (1, 2, 3, 4)

Age (1, 2, 3)

I understand how to use proc Genmod to estimate betas, but I'm trying to estimate crude rate ratios for each variable and then the rate ratios for an adjusted model with multiple variables.

What would estimate statement look like comparing males to females? Season 3 to the rest?

For the adjusted model, would I just be better off calculating it using the betas?

Finally, regarding how to interpret the parameter estimates provided in the absence of estimate statements-- If I have output that gives me a parameter estimate of 0.1493 (95% CI: -0.0823, 0.3809) for sex=0, and I exponentiate the estimate, is it correct to say "females are expected to have 1.41 (95% CI: 0.83, 2.40) more events than males per person-year"?

Thanks in advance for the guidance!

Re: Writing Estimate Statements and Interpreting Poisson Output

StatDave — Wed, 13 Nov 2019 19:46:37 GMT

For simple comparisons like this, you should use the LSMEANS statement rather than the more complicated ESTIMATE statement which requires determining appropriate contrast coefficients which is an error prone task for most people. See this note which discusses and illustrates using the LSMEANS statement for estimating rates and rate ratios.

topic Writing Estimate Statements and Interpreting Poisson Output in Statistical Procedures

Writing Estimate Statements and Interpreting Poisson Output

Re: Writing Estimate Statements and Interpreting Poisson Output