Data Thinning to Avoid Double Dipping in Statistical Analysis - Lecture 1

Explore the concept of "double dipping" in statistical analysis and learn about innovative approaches to address this issue in a lecture from the Models, Inference and Algorithms series at the Broad Institute. Dive into Lucy Gao's presentation on data thinning techniques, focusing on Poisson thinning and its generalizations, which offer solutions for avoiding double dipping in unsupervised settings, particularly in single-cell RNA sequencing data analysis. Gain insights into the challenges of sample splitting and discover how data thinning can be applied across various distributions and problem domains. Additionally, benefit from Yiqun Chen's primer on testing data-driven hypotheses post-clustering, addressing the statistical validity concerns in biomedical research when generating and testing hypotheses from the same dataset. Learn about a conditional selective approach for testing differences in means between clusters obtained through hierarchical and k-means clustering, with applications in single-cell RNA-sequencing analyses.