![](https://www.maynoothuniversity.ie/sites/default/files/styles/ratio_2_3/public/assets/images/HamiltonInstitute_Singleheader1_MaynoothUniversity3.png?itok=PeA-ggHn)
Virtual participation: Zoom details available here
Speaker: Professor Claudia Tarantola, University of Pavia
Title: "Simplicity in Action: Interpreting Effects in Binary and Ordinal Data Analysis Models"
Abstract: Modelling binary or ordinal response variables with nonlinear link functions often leads to effect parameters that are not as simple to interpret as slopes and correlations for ordinary linear 42 regression.
This presentation addresses this issue by introducing probability-based measures tailored for such models. Specifically, we explore the utilization of identity and log link functions to generate straightforward effect measures when describing the impact of an explanatory variable on a binary response, even while accounting for other covariates. In cases where these link functions are unsuitable, we show how one can construct analogous effect measures using standard models like logistic regression.
Additionally, we delve into effect measures for ordinal response variables, showcasing both the standard cumulative link models based on the proportional odds assumption and the recent extension of the Combination of Uncertainty and Preference models. This includes the introduction of mixture models, which effectively accommodate uncertainty within rating systems. The talk concludes with a discussion on how to construct effect measures for generalized additive models.
This talk is based on recent works with Alan Agresti, Maria Iannario and Roberta Varriale.