Neg Bin Regres Output

[#1]

Hello all,

I am a student working on a research project and have a question/clarification regarding how to interpret the Fit Model output for a negative binomial regression. I am trying to predict a count variable (Variable A = how much time it took for an event to happen) based on sex (M vs F). I am using a negative binomial regression because my count variable is not normally distributed and my data is over-dispersed (I think this is the right approach but let me know your thoughts on this as well). My plan is to repeat this with different predictors and then include all the ones that are significant in a single model with multiple predictors.

When interpreting the report, my p-value for the Sex[F-M] term is significant, showing that sex is a predictor of Variable A. The generalized RSquare is very low so I know it's not a great fit.

--> I am trying to understand how to say that for a given sex (i.e. Females), Variable A is XX % higher (or lower). How do I do that?

--> How do I interpret the Estimate?

In the same report, I am also trying to look at Incidence Rate Ratios. If,

Level 1 = F, Level 2 = M, IRR = 1.303

--> Does that mean that for F compared to M, there is a 30% increase in Variable A?

Thanks for all the help and sorry if my explanation is not clear! I am just getting started with statistics...

[#2]

The answers to most of your questions are documented here. Additionally,

• "I am using a negative binomial regression" seems reasonable but I am not sure that I would categorize your response as 'count' data. You are measuring time, not counting items. But the choice of the distribution depends on the response, so go with it.
• "My plan is to repeat this with different predictors and then include all the ones that are significant in a single model with multiple predictors." I interpret your plan to model each predictor one at a time the same way as you modeled with Sex already. I advise you to instead enter all the predictors in the same model and then select the active terms / chosen model in the usual way.
• "how to say that for a given sex (i.e. Females), Variable A is XX % higher (or lower)" is probably best handled with the Prediction Profiler, which is available from the platform red triangle menu.
• "How do I interpret the Estimate?" is answered in this chapter of JMP Help.
• "Does that mean that for F compared to M, there is a 30% increase in Variable A?" is answered on the page I suggested first above.