classical frequentist and Bayesians are not the only schools of statistics useful for animal models

http://www.quantumforest.com/2011/11/coming-out-of-the-bayesian-closet/#comments


I would also like to point out that classical frequentist and Bayesians are not theonly schools of statistics useful for animal models.

Classical frequentist inference uses marginal likelihood (random effects integratedout) that includes fixed parameters and only the observations are treated asrandom.
Bayesian inference is a probabilistic framework that combines likelihood and prior information, and treats all parameters and observations as random.
A third important school is the one based on the Extended Likelihood Principle. Assuming simple statistical principles Bjørnstad (1996) showed that all information in the data about the random and fixed effects is included in a joint likelihood including three components: fixed parameters, unobserved random effects, and observations as random. Lee and Nelder's (1996) h-likelihood is an implementation of the Extended Likelihood Principle. The hglm package (that I have developed together with colleagues) is based on h-likelihood theory and allows fitting of animal models.

hglm: Hierarchical Generalized Linear Models