## Problem with Poisson Model for Rates

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

IUKC
Posts: 2
Joined: Sat Aug 04, 2012 4:14 pm

### Problem with Poisson Model for Rates

Dear all,

I have recently been given a piece of SPSS-based coursework, and with the deadline coming into view I am stuck on the data entry section. I am using SPSS version 19.

I have been given two tables of data: Table 1 presents the number of cases of insulin-dependent diabetes mellitus (IDDM) by sex, age and geographical area.
Table 2 presents the estimated population at risk by sex, age and geographical area.
There are 6 areas and 3 age groups (0-4, 5-9 , 10-17), and both cases and populations are counted separately for Males and Females.

I am meant to enter the data appropriately into SPSS and then select a suitable Poisson model for the IDDM rates given age, sex and area. To do this, I would have to calculate a constant parameter in order to create the initial constant rates model, and then expand on it.

Unfortunately, the only Poisson model for rates I have as a guide has gender-neutral cases and population at risk – not separate for males and females like in this case. As a result, I am not sure how to input the data into SPSS:
if I put in separate male and female data, I will get two different constants which will require a model each (I have to limit myself to one); but if I merge the male and female data in order to get a single constant, the model will no longer be able to highlight differences between male and female rates, making it more imprecise.

Does anyone have suggestions on how to solve this dilemma?

Penguin_Knight
Posts: 473
Joined: Thu Apr 05, 2012 5:58 pm

### Re: Problem with Poisson Model for Rates

I don't understand what the problem is. If gender is in the regression model, then you should be able to obtain the predicted rate for both females and males. How would it be "gender-neutral"?
IUKC
Posts: 2
Joined: Sat Aug 04, 2012 4:14 pm

### Re: Problem with Poisson Model for Rates

Dear penguin_knight,

Thank you for your response! I ended up understanding what I had to do for data entry whilst looking at an example's Data Editor, and I've now run regressions for all possible models (7 in all, I'm not sure whether I should include interactions at the present time) and think I've found the one that fits the data best.

Still have some doubts over what method is best to compare models in order to find the one with the best fit though.

IUKC

### Who is online

Users browsing this forum: No registered users and 2 guests