# Thread: Effect Size in Poisson Reg with a Binary Predictor

1. ## Effect Size in Poisson Reg with a Binary Predictor

Well, I don't regularly come across count variables as dependent variables. What is the best effect size to report for a Poisson reg model with a binary predictor?

So say I got the following output for my binary independent variable:

beta-hat: 0.0093, SE: 0.0873, (95% CI -0.1618, 0.1805), p-value 0.9149.

I have seen the following definition:

"This is the estimated Poisson regression coefficient comparing group 1 to group 0, given the other variables are held constant in the model. The difference in the logs of expected counts is expected to be 0.0093 unit higher for group 1 compared to 0, while holding the other variables constant in the model. This is the estimated Poisson regression coefficient comparing groups, given the other variables are held constant in the model. The difference in the logs of expected counts is expected to be 0.0093 unit higher for group 1 compared to group 0, while holding the other variables constant in the model.

Should I be exponentiating the value and if so what would be my interpretation, just drop the log part in the above statement??

Thx

2. ## Re: Effect Size in Poisson Reg with a Binary Predictor

I am thinking I can get the rate ratio, but need to look this up.

3. ## Re: Effect Size in Poisson Reg with a Binary Predictor

So my count data is not a rate or easily converted to a rate. It is a number (count) of variable X. So is there a classier way to say the following or is this what you all use?

The predicted mean count of Y for Group = 1 is 0.99 times that of Group = 2. I forgot the negative signs for flipping the groups in post #1. And if I had to give this a name would it be the rate ratio, given the X variable is not a rate but a count?

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts