Abstract:
|
In many medical researches, measurements obtained from paired organs (e.g., eyes or ears) of an unit are generally highly correlated. It is very important to account for the intraclass correlation on statistical inferences, since ignoring the intraclass correlation between paired measurements may yield biased inferences. In addition, it is commonly needed to consider simultaneous comparison of proportions of success between a single control group and multiple treatment groups in randomized clinical trials. In this research, we constructed simultaneous confidence intervals (SCIs) for odds ratio in a many-to-one comparison framework under such correlated paired binary data. Four different methods are applied to construct simutaneous confidence interval for odds ratio with Dunnett-like or Bonferroni multiple adjustment. The empirical coverage probabilities and mean interval widths of the SCIs from resulting methods are compared through a Monte Carlo simulation study to evaluate their performance. A real work example is included to illustrate the usage of the resulting methods.
|