How to compute risk ratio in logistic regression?

[Tom Hsiung]

For logistic regression we have:

• Parameter 1: P (probability of event [event present =1])
  P = 1 / {1 + exp(-logit[P])}
• Parameter 2: Odd
  Odd = exp(logitP)
• Parameter 3: logitP
  LogitP = 𝛽0 + 𝛽1 x 1 + … + 𝛽k x k

Odd ratio = Odds ratio = exp[𝛽1(x 0 - x 1) + 𝛽2(x 0 - x 1) + … + 𝛽k(x 0 - x 1)] = exp[𝛽1(x 0 - x 1)] * exp[𝛽2(x 0 - x 1)] * … * exp[𝛽k(x 0 - x 1)]

How to compute the risk ratio?

Tom

[Marcos Almeida]

For this, you are supposed to use a generalized linear model with a binomial family and a log link, and use the option to get exponentiated coefficients.

[Rich Goldstein]

or use -poisson- regression

[Tom Hsiung]

For this, you are supposed to use a generalized linear model with a binomial family and a log link, and use the option to get exponentiated coefficients.

I have to use logit regression because the response variable is binary (0 or 1)

Tom

[Marcos Almeida]

Unfortunately, you didn’t take a close look at the model I shared in #2.

No doubt, you can use the model in #2 for a binary (0 or 1) depvar.

Please take some time to grasp the core knowledge related to the Issue.