Searching for a source of difference in undirected graphical models for count data: an empirical study

Abstract: A study is presented for exploring the possibility of applying the Source set approach, developed under the assumption of normality, to count data, after data transformation. Some explanations about the source set approach, data transformations and the simulation setting are provided. The suggestion is given that the deviance-based or quantile randomized residuals could provide a better basis for data transformation when coupled with source set analysis, along with standard trasformations such as log transformation or square root transformation.

Key words: Undirected graphical models, Source set, count data, negative binomial distribution, RNA-Seq data

Recommended citation: Agostinis, F., Chiogna, M., Djordjilovic, V., Pianesi, L. and Romualdi, C. (2021). "Paper Title Number 1." Book of short papers. SIS - Società Italiana di Statistica, 689-694.
Download Paper