Sample Size Consideration on Search for Correlative Models from High-Throughput Screening Data

• Activity Number: 237
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Xian-Jin Xie*+ and Yijie Liao and Jason Yan
• Companies: The University of Texas Southwestern Medical Center at Dallas and Southern Methodist University and The University of Texas Southwestern Medical Center at Dallas
• Keywords: high throughput screening ; sample size ; model search

Abstract

We consider the limitations on search for correlative models from high throughput screening data where the number of potential predictors far exceeds the number of the sample replicates. Overfitting as well as intrinsic low power of detecting true associations from such data should always be taken into account when reporting the results. Under certain assumptions, we derive the calculation of the required sample size for detecting the true model from high throughput data. The sample size required is expressed as a function of the location of the rank of the true model among the best models derived from exhaustive search algorithm. Simulation studies are performed to evaluate the results. Conditions for our assumptions and possible further relaxation of the assumptions are discussed.