In 1994, Lee and Nelder introduced the “h” type of these functions, which remain controversial for their handling of unobservables. These functions are often approximated via Gauss–Hermite quadrature in GLMMs. Two transformed examples of these functions are used to find McFadden’s pseudo-R-squared in logistic regression. A theoretical approach named for these functions makes use of a law that was stated by Ian Hacking, and concerns ratios of two of them denoted by lambda. The EM algorithm works on these functions or on the posterior probability. These functions, which are commonly log-transformed, represent the joint probability density of observing data, given a set of parameters. For 10 points, name these functions whose “maximum” value names a common paradigm of parameter estimation, and which are the “L” in MLE. ■END■
ANSWER: likelihood functions [accept maximum likelihood estimation, log-likelihood functions, likelihood ratios, log-likelihood ratios, likelihoodism, likelihoodist, law of likelihood, or h-likelihood functions; prompt on MLE until read by asking “what does that stand for?”]
<Other Science>
= Average correct buzz position