|dc.description.abstract||The double-sampling paradigm, which has become an important part of the epidemiological designs, includes two stages. First, individuals are classified into groups by disease and exposure levels using a fallible test, and second, some individuals are classified into a subset using a ``gold standard" test. The parameter of interest in our study is the odds ratio as an association between disease level and exposure level. Here we compare four confidence intervals for the odds ratio under the assumption of differential or non-differential misclassification. More specifically, we compare the coverage and interval widths of the Wald, score, profile likelihood, and approximate integrated likelihood intervals with different specificity and sensitivity values, as well as different sample sizes and odds ratios for the case-control clinical studies. Our investigations implies the consistent superiority of the approximate integrated confidence interval.
Also, we eliminate the effect of several parameters on a bioequivalence testing procedure that plays an important role in the development of generic drugs. The current FDA criteria is not flexible with respect to highly variable drugs, and this characteristic has caused many good drugs to be rejected. Most often in the literature, we find studies examining the sample size or the within-subject variability as the main factors affecting the outcome of a bioequivalence test. Frequently, pharmaceutical companies have tried to convince the FDA that their product would meet the bioequivalence criteria just by increasing the sample size. Here we examine the effect of the between-subject variability as well as the effect of the mean ratio difference between the test and reference formulations. We use a Monte Carlo simulation to draw conclusions based on the importance of these two sources of variability and to show that simply increasing the sample size is insufficient to meet the bioequivalence criteria.||en_US