Likelihood-Based Statistical Inference with ELiISA Package

Esa Uusipaikka
University of Turku

ELiISA (Efficient Likelihood-Based Interactive Statistical Analysis) is a package for calculation of maximum likelihood estimates of parameters, likelihood ratio tests of hypotheses about parameters, profile likelihood-based confidence intervals for real-valued parameter functions of interest in various statistical models, such as sampling, univariate/multivariate linear/nonlinear regression, logistics regression, Poisson regression, multinomial regression, log-linear, generalized linear models, etc.

ELiISA handles the distributions of statistical models symbolically and provides to the user access to the cumulative distribution, density, log-likelihood, score, observed information functions as functions of parameters and observations. It gives access to the generating functions, moments, cumulants, and expected informations as functions of parameters. This symbolic handling is necessary for the calculation of certain adjustments in approximate likelihood-based statistical inference. ELiISA uses a new method for calculation of profile likelihood-based confidence intervals for general parameter functions of interest in general parametric statistical models.