Count data regressions

Question 1

Poisson distribution

Answer 1

Distribution only valid for non-negative integer values (y has no decimal) & has a right-skewed shape

Question 2

Model offset term

Answer 2

Rate of occurrence over time period t: log(λ/t)

Model with offset: log λ = log(t) + β0 + β1x1 + β2x2

Example: if interested in death rate of children for all women, outcome is a function of total # children a woman has had

Question 3

Negative binomial regression model

Answer 3

Potentially use if overdispersion in Poisson regression.

Question 4

Zero-inflated Poisson model

Answer 4

Fits Poisson model while modeling probability of the count=0 using a binomial model

Question 5

Assumptions of Poisson regression

Answer 5

1) Log transformed outcomes linearly related to predictors

2) COME BACK