Testing and analytical procedures for laboratory studies involving nonresponders during a limited observation period: an illustration using male sexual behavior in rats
In many laboratory studies, a subpopulation of subjects fails to exhibit the response under investigation during the period of observation. For example, within any population of male rats, there is significant variation in the expression of sexual behavior in the presence of a receptive female. Some males may never display the full sequence of behaviors leading to ejaculation within the typical time frame of the testing session, with the resulting lack of behavioral response presenting problems in the analysis of the data. Conventional strategies range from screening such males from the study or dropping them from the analysis to constructing new variables based on estimates from existing parameters or increasing the length of the test session to capture sexual responses in a greater portion of males. Herein, we present an alternative strategy for analyzing data where outcomes are absent due to the limited observation period. Survival regression analysis enables inclusion of all subjects in the analysis whether or not they have shown the behavior of interest. Use of such a strategy not only has potential to reveal new results but also guards against bias from excluding nonresponders from the study or dropping more males from one experimental condition than another. Furthermore, this procedure can be helpful in generating the conditional probability (increase, decrease, or constant) of the response with the passage of time based on the hazard function and in estimating parameters for establishing an optimal behavioral test length for future studies.
Rowland, D. L., & Thornton, J. A. (2001). Testing and analytical procedures for laboratory studies involving nonresponders during a limited observation period: an illustration using male sexual behavior in rats. Pharmacology Biochemistry and Behavior, 68(3), 403–409. doi:10.1016/S0091-3057(00)00473-1