Correlation of Sequential Binomial Variables and Its Application to Multiple Testing

• Activity Number: 617 - Testing
• Type: Contributed
• Date/Time: Thursday, August 1, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Lin Fei* and Changchun Xie
• Companies: Cincinnati Children's Hospital Medical Center and University of Cincinnati
• Keywords: binomial variable; conditional testing; multiplicity adjustment

Abstract

We study a sequential testing of conditional binomial probabilities to assess incremental rates of success of several interventions. This design is more efficient than parallel procedures. In order to make use of dependent relationship among test statistics, thus to control family-wise error rate, we obtained an recursive procedure to calculate correlations for the sequence of observed binomial random variates. They were derived through conditioning on the natural filtration associated to the test statistics.