Negative Binomial Regression — reg

Fits a negative binomial regression model (for over-dispersed counts) via MASS::glm.nb().

Usage

reg_negbin(formula, data, step = FALSE, ...)

Arguments

formula: A formula, e.g. count ~ x1 + x2.
data: A data frame.
step: Logical. Always FALSE for negative binomial.
...: Additional arguments passed to MASS::glm.nb().

Value

A named list analogous to reg_poisson(), plus:

theta: Estimated dispersion parameter for the negative binomial distribution.
residuals: Tibble of observed, fitted, Pearson residual.

Details

Note: The glm.nb family does not support stepwise selection. Set step = FALSE.

Unlike reg_poisson(), the negative binomial model handles overdispersion through the estimated dispersion parameter theta (fitted by MASS::glm.nb()): a small theta indicates strong overdispersion, while theta growing large approaches the Poisson model.

Examples

set.seed(42)
x1 = rnorm(100)
mu = exp(1 + 0.3*x1)
y = rnbinom(100, size = 1, mu = mu)
res = reg_negbin(y ~ x1, data = data.frame(y, x1))
res$coefficient
#> # A tibble: 2 × 7
#>   term        estimate std.error statistic  p.value conf.low conf.high
#>   <chr>          <dbl>     <dbl>     <dbl>    <dbl>    <dbl>     <dbl>
#> 1 (Intercept)    0.907     0.117      7.74 1.01e-14   0.674      1.14 
#> 2 x1             0.322     0.116      2.78 5.52e- 3   0.0918     0.553