Multiple Comparisons скачать в хорошем качестве

Multiple Comparisons

Presenter: Brent Logan, PhD, Professor, Division of Biostatistics, Medical College of Wisconsin Many studies are conducted with multiple hypotheses being tested. These can include considering multiple outcome variables, comparing more than two groups with one another, and examining thousands of genes for disease association. If the multiple testing is not accounted for, the likelihood will increase that a spurious or false significant finding appears in the data. This talk will examine the impact of multiple testing and when it is appropriate to adjust. We will discuss several techniques for adjusting the analysis for multiple comparisons. Finally, we will focus on the large multiplicity problem, where thousands of tests are being performed in a single study, and review procedures specific to this setting.