Principal investigator: Aiyi Liu, Ph.D.
Sequential methods have become a common practice in biomedical research as a result of reasons such as ethical, administrative, and economic concerns. In testing statistical hypotheses, the methods--known as group sequential tests or repeated significance tests--allow hypotheses to be tested repeatedly at several time points after a group of observations have been completed. In designing a biomedical study, the sequential concepts allow the study design to be modified, a process known as adaptive procedure; for example, re-estimating the sample size needed to achieve the desired power, or changing the allocation ratio of cases to controls based on analysis of interim data from midcourse of the study. Sequential selection procedures, often in the form of two-stage selections, have been used in analysis of high-dimensional data as a means of dimension-reduction. For example, in genome-wide association studies, two-stage selection has been widely used to select diseased-associated single nucleotide polymorphisms. Sequential methods have only recently been investigated in the area of diagnostic medicine and evaluation of measurement errors of diagnostic biomarkers.

The sequential methods pose great challenges to statisticians. These challenges include not only finding appropriate methods for determining critical values at each interim test to control error rates, such as false positive/negative rates, but also developing efficient inferential procedures for secondary analysis, such as estimation, confidence intervals, P-values, and secondary endpoints. Because even the data-based random stopping of the sampling process introduces bias to conventional procedures that are appropriate in a non-sequential (fixed-size) setting, statisticians are very interested in this area of research.

DESPR research in this area has been focused on two directions: first, finding proper sequential procedures suitable for various biomedical studies, such as evaluation of a biomarker’s measurement error, testing hypotheses concerning the sensitivity, specificity, or the receiver operating characteristic curves of diagnostic biomarkers; and secondly, developing efficient statistical procedures for data analysis after the stopping of a sequential procedure.

DESPR Collaborators

Selected Publications

Liu A, Wu CQ, Yu KF, & Yuan V. (2007). Estimation following a multivariate group sequential test. Journal of Multivariate Analysis, 98, 505-516.

Liu A, Hall WJ, Yu KF, & Wu CQ. (2006). Estimation following a group sequential test for distributions in the one-parameter exponential family. Statistica Sinica, 16, 165-181.

Liu A, Wu CQ, Yu KF, & Gehan E. (2005). Supplementary analysis of probabilities at the termination of a group sequential phase II trial. Statistics in Medicine 24, 1009-1027. [Abstract]

Troendle JF, Liu A, Wu CQ, & Yu KF. (2005). Sequential testing for efficacy in clinical trials with non-transient effects. Statistics in Medicine, 24(21):3239-3250. [Abstract]