Chiba, Yasutaka and Banba, Ikuko and Azuma, Kenichi (2010) The principal strata effect for post-randomization selection bias : application to a prostate cancer prevention trial. Biomedical statistics and clinical epidemiology , 4 (1). pp. 47-53. ISSN 1973-2430
PDF
portiere.pdf - Published Version Restricted to Repository staff only Download (590kB) | Request a copy |
Abstract
Objectives. An approach that deals with post-randomization selection bias is to consider the principal strata effect (PSE). In this paper, the concept of the PSE is introduced, aimed at researchers. The practicality of a sensitivity analysis method that has recently been developed is also investigated. Although a formula for the method had been presented previously, its practicability with real data has not been discussed for general cases. Methods. The PSE is introduced using a simple hypothetical example. Whether the sensitivity analysis method is practical is examined by applying the process to an actual prostate cancer prevention trial. Results. The sensitivity analysis could be readily conducted, and the results could be displayed graphically using Monte Carlo sensitivity analysis. Conclusions. The sensitivity analysis requires prior distributions of sensitivity parameters. Thus, it will be useful for diseasespecific experts, who have sufficient clinical knowledge, and statisticians, who can conduct the sensitivity analysis, to work closely together for the sensitivity analysis.
Actions (login required)
View Item |