My sample is an environmental sample. Is there anybody who can help me with this?
Hello Shahadat,
Have a look at this link, which walks you through the process using R:
https://rcompanion.org/handbook/J_02.html
Good luck with your work.
The beta regression is a widely known statistical model when the response (or the dependent) variable has the form of fractions or percentages. In most of the situations in beta regression, the explanatory variables are related to each other which is commonly known as the multicollinearity problem.
I also recommend David's link. It explains when you can use logistic regression or Poisson regression instead. However, if you are bound with beta regression, an alternative to using the betareg package is to fit e GLM with family = "quasipoisson". The results (coefficient estimates, standard errors, p-values etc.) will be (essentially) identical to that of betareg, the only difference is that you don't get the maximum likelihood (and hence also no information criteria) of the fitted model. If you don't need that, the quasipoisson model is simple and robust.
In addtion to Jochen and Davids answer, another option would be the gamlss package for R (https://www.gamlss.com/distributions/) it also contains a function to perform beta regression (and others, such as zero-inflated and zero-one-inflate). It is relatively versatile, yet the expression are slightly different than the standard R base notations therefore you might run in some unexpected errors. Yet, considering the given options above there is enought to choose from.
- Badges
- Science topic
- Similar topics
- Biological Science
- Biota
- Eukaryota
- Fungi