I have overdispersed count response data "number of children". I want to fit the negative binomial regression model. In the process, I faced a problem in fixing the value of the dispersion parameter "ancillary parameter". Thus, my question is how can I fix the dispersion parameter in the Negative Binomial regression model? is there any recommended cut-value to use?
There is a possibility in R:
See https://cran.r-project.org/web/packages/pscl/vignettes/countreg.pdf, page 5:
Package MASS (Venables and Ripley 2002) provides the family function negative.binomial()that can directly be plugged into glm()provided the argument theta is specified. One appli-cation would be the geometric model, the special case where θ=1, which can consequentlybe fitted in R by setting family = negative.binomial(theta = 1) in the glm() call.
So attach MASS and use glm(..., family=negative.binomial(theta=fixed_value))
The variance of a negative binomial distribution is a function of its mean and has an additional parameter, k, called the dispersion parameter. Say our count is random variable Y from a negative binomial distribution, then the variance of Y is. var(Y)=μ+μ^2/k.
>I want to fit the negative binomial regression model....
And what about Your data?
They also want to be distibuted according negative binomial distribution?
======
Sometimes changing the model helps. For examle to zero-inflated poisson... or zero-truncated poisson...
Look at distribution at first
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis