Class 11 (Module II)

27 Abril 2022, 15:00 • Jorge Filipe Campinos Landerset Cadima

[GLM Slides 80-152 (except 126, 127)] [Video, Part 1] [ Video Part 2] Asymptotic inference theory for GLMs, based on the asymptotic properties of Maximum Likelihood estimators: confidence intervals and hypothesis tests for individual parameters or linear combinations of the parameters. Commands for inference in R and an example. GLMs with Poisson response variables and Log-linear models. An example with emergence data (GLM Exercise 5). Deviance in GLMs as a measure of goodness of fit: definition and interpretation. Deviance formulas for Poisson response models. Likelihood Ratio Tests (LRT) or Wilks' test to compare nested models, or as a goodness-of-fit test, for models that do not require the estimation of the dispersion parameter phi. Again the emergence data example (GLM Ex.5). ANCOVA-type GLMs: Exercises 1 and 10 (tobacco budworm data). Akaike Information Criteria in GLMs. The issue of an unknown dispersion parameter phi: the need for simplifying assumptions that make estimation possible. Deviance and Scaled Deviance. Residuals: Pearson Residuals and Deviance Residuals: definitions and interpretation for specific contexts. The Generalised Pearson Statistic and its use in estimating the dispersion parameter phi. AS brief note on the use of Log-linear models in the study of contingency tables.
Nota: Esta aula foi leccionada na segunda-feira, dia 9.5.22, das 9h45 às 12h25, na Sala S1, simultaneamente presencial e por zoom.