Estimates overdispersion in multinomial models — estimate

This is an adaptation of code found spread out in the package mclogit. The methods for estimating overdispersion are described in Afroz 2018. Summarized in page 54 and 55.

Usage

estimate_overdispersion(
  y,
  fitted_values,
  coefs,
  method = "Afroz",
  number_samples = NULL,
  weight = NULL
)

Arguments

y: The vector of multinomial observations.
fitted_values: The fitted mean values.
coefs: The Coefficients or the number of coefficients.
method: The method of estimating overdispersion. One of "Afroz", "Pearson", "Fletcher", or "Deviance". "Afroz" is recommended.
number_samples: The number of samples that were fitted. Usually the number of rows in the matrix fitted (y). Only necessary if providing a flattened y.
weight: TODO

Value

The estimation for overdispersion.