Statistical principles in neurointervention part 2. Multivariable analysis: generalized linear models, modification, confounding, and mediation.

Abstract

This two part series on statistical principles in neurointervention offers a comprehensive foundation for neurointerventionalists to engage with both fundamental and advanced statistical principles. This series aims to equip neurointerventionalists with essential statistical knowledge for critically reviewing literature and conducting methodologically sound research. Part one of this series covered fundamental concepts such as frequentism, study types, data types, summarization, visualization, hypothesis testing, and univariable analysis. This review is the second part of the series and covers advanced statistical concepts such as inference versus prediction, multivariable analysis, choice of covariates, confounding, mediation, modification, and generalized linear models. Together, these papers create a cohesive framework, allowing practitioners to critically evaluate research and apply rigorous statistical methods to their own studies.