- Location
- 1 King's College Circle, MS4171
- Series/Type
- Biostatistics Seminar, DLSPH Event
- Format
- In-Person
- Dates
- April 16, 2024 from 3:10pm to 4:00pm
The Biostatistics Seminar Series presents:
“Assessing Heteroscedasticity and Model Misspecification in GWAS for Gene-Environment Interaction” by Dr. Konstantin Shestopaloff, University of Aberdeen, UK
Abstract: This presentation will go over techniques to account for heteroscedasticity when evaluating the significance of gene-environment interactions in GWAS studies. The presented work was done as part of an analysis for a recently published paper assessing the role of genetics on treatment efficacy for autoimmune diseases. As part of the analysis, the inference model used to assess gene-environment interactions in a set of clinical trial data was assessed for misspecification and heteroscedasticity using a simulation study. Results indicated sparsity as the main cause of inflated type 1 error when using heteroscedasticity adjusted variances. A number of adjustments were made to minimize false positive rate while retaining model power.