search

atahk / pscl

Political Science Computational Laboratory
65 stars 14 forks source link

Estimation of dispersion parameter (theta) of NB #14

Open WisamSaleem opened 4 years ago

WisamSaleem commented 4 years ago

Dear Achim. Hi. I have two questions:

  1. Using the function zeroinfl() with distribution is "negbin", could you please help me to find out which version of NB has been used there? Is it NB2 with unknown dispersion via MASS::glm.nb() for example? Or is it from some other NB function with fixed theta like MASS::negative.binomial()?

  2. As I implemented the examples in the tutorial, I found out that theta is estimated in zeroinfl() with distribution is "negbin", then I tried to check that theta, with theta from MASS::glm.nb() for the same example. They are close to each other, but not exactly the same! Do you have any interpretation for this? Many thanks

Wisam

MarcRieraDominguez commented 1 year ago

Hi! From what I've explored, I think pscl::hurdle() fits type 2 negative binomial (quadratic relationship between mean and variance). See a reproducible example showing this in this Stack overflow question.