Abstract
Statisticians working in biological investigatory trials are quite often faced with inferences obtained in long-range experiments which do not fit within the general framework of time-indexed designed studies. The initial response of patients in long-term studies is usually depicted within the first six months of the trial; the remaining part of the trial being confined to patient monitor of dropouts, survival rates and adverse effects. The objective of the present paper is to describe two computerized profiling methods developed by the present author for maximizing contrastive differences between medications.
The first method analyzes the consistency of successive observations on a patient-by-patient basis, yielding cumulative increment scores which is used as a decision rule for a time-indexed response. The method is reminiscent of sequential medical trial procedures applied to data from a given patient. Using this procedure, patients are categorized as “responders” or “non-responders” depending upon the profile of successive observations which they elicit during the trial.
The second method utilizes superimposed distribution profiles of each medication under comparison. Profiles generated at various cut-off points during the study enable the utilization of more stringent indices than means and variances of the observations. Constriction in data variability is obtained by utilization of non-parametric trichotomization. Double-blinding of comparative medications may still be kept intact: yet the investigators have the flexibility of monitoring emerging contrastive differences as the trial proceeds.
Get full access to this article
View all access options for this article.