Abstract:
|
Desirability functions have been widely used in the industrial and manufacturing sectors to measure quality of a product. A desirability score is a continuous measure between zero and one, with large values representing greater desirability. An individual desirability function is constructed for each component that contributes to the quality of a product. The desirability functions for the individual components are then combined with weights reflecting relative importance to product quality to form an overall desirability score.
We use desirability functions to measure the appeal of many optimal adaptive design strategies based on statistical and non-statistical components that contribute to the quality of the trial. Specifically, we construct individual desirability functions for a library of these components that include total sample size, power, accidental bias, chronological bias, expected number of failures, financial cost, and anticipated timeframe for completion. The overall desirability for each strategy is then computed. This allows the user to rank a family of clinical trial designs and select the most relevant and efficient design for the trial's various objectives.
|