Mulitiple Comparisons Procedures in the Presence of Missing Data

Mr. Joseph Kosler
Assistant Professor
Mathematics Department

Key Terms:  multiple comparisons procedure, missing data, multiple imputation, balanced design, spherical model for covariance, two-way mixed-effects repeated-measures model with interaction.
Missing data may render a multiple comparisons procedure (MCP) ineffective.  We will discuss methods for performing multiple comparisons on treatment effects in the presence of missing data.  In particular, we will discuss the case of the two-way mixed-effects repeated-measures model with interaction term.  MCPs under this model challenge the notion of a balanced design, and emphasize the need for imputation procedures to preserve the underlying covariance structure (spherical model).  Ultimately, we will discuss the method of multiple imputation as both effective and necessary for performing MCPs under a two-way mixed model with interaction.

Back to the Colloquia Calendar