Principal speaker
Associate Professor Sama Low-Choy
Presuming that you understand the purpose and The Meaning of Logistic regression (3 episodes: model, explainability, predictions), this series of 4 episodes walks you through the process of doing logistic regression. However, this is *not* a software-based training - we refer to the kinds of outputs that you may obtain from any statistical package and/or virtual laboratory that can conduct logistic regression (e.g. R, Stata, SPSS, Statistica, SAS, etc). In the context of logistic regression, we show the benefits of thinking about what you are doing, and conversely, the dangers of not thinking whilst doing.
Credible effects, building on prior information: A Bayesian approach introduces a prior model for effect sizes, which subtly changes their interpretation, whilst also expanding the possibilities for statistical modelling (w.r.t. sample size, non-random samples, multiple comparisons). We show the necessity of considering a noninformative prior reflecting ignorance, and the possibilities for informative priors where previous information is available.
Format: These 4 'Stat-a-along' walk through thinking whilst doing logistic regression.
Intended audience: Those with no experience in regression, wishing to walk through the process of doing logistic regression, whilst still thinking about the outputs and what they mean. Those with some conceptual understanding of what logistic regression means, wishing to go to the next level, and put this into practice.
Preparation: We strongly suggest that you revise the concept of a function and of logarithms, revising topics of high school mathematics, presented sometime during Grades 9-11, e.g. refer to the friendly You-tubes of Mr Wu, an Australian maths teacher who won national acclaim.