I used the Conway-Maxwell-Poisson distribution for my underdispersed count data. Now I'm interested in compare the treatment but I don't found the script for it.
For the others dsitributions (Poisson, binomial Negative) there is the emmeans packege , that works very well.
What I would use to compare treatments?
Hi Nadia Diovisalvi
You may want to contact the maintainer of COMPoissonReg package, Andrew Raim , or the maintainer of CompGLM, Jeffrey Pollock .
It seems like emmeans does not yet support a COM-Poisson object or a compGLM object.
These packages, however, provide support for hypothesis testing via equitest or parametric bootstrap. For example, see https://rdrr.io/cran/COMPoissonReg/man/COMPoissonReg-package.html
If nothing else helps, you can re-fit the model using formulas that have coefficients for some of the contrasts you are interested in. You can collect the p-values (from the likelihood-ratio tests) for these coefficients and finally correct the collected p-values for multiple testing (e.g. by Holm's procedure).
If you have a factor "treatment" and you want all pair-wise comparisong of the different treatments, you can relevel the factor for each of the levels and refit the model using the releveled factor. Each time you will get coefficients comparing all levels to the current reference level.
https://cran.r-project.org/web/packages/COMPoissonReg/COMPoissonReg.pdf
contact:
Kimberly F. Sellers ([email protected])
- Badges
- Science topic