Abstract
A clinical trial model is considered in which k > 2 treatmentsare compared and treatment allocation is data-dependent. A sequential procedure for determining the best treatment is investigated that is a natural generalization of the test for two treatments studied by Robbins and Siegmund (1974). It is shown by extensive simulation that the error probability for the procedure is insensitive to the data-dependent allocation rule used. The estimation formulae of Coad (1994) are shown to give good approximations to the bias and variance of estimators of treatment differences.
| Original language | English |
|---|---|
| Title of host publication | Adaptive Designs |
| Subtitle of host publication | Papers from the Joint AMS-IMS-SIAM Summer Conference held at Mt. Holyoke College, South Hadley, MA, July 1992 |
| Editors | William F. Rosenberger, Nancy Flournoy |
| Place of Publication | Hayward |
| Publisher | Institute of Mathematical Statistics |
| Pages | 95-109 |
| Number of pages | 15 |
| Volume | 25 |
| ISBN (Print) | 0940600366 |
| Publication status | Published - 1 Jan 1995 |
Keywords
- Brownian motion
- clinical trials
- data-dependent treatment allocation
- error probabilities
- estimation
- Gittins index
Fingerprint
Dive into the research topics of 'Sequential allocation involving several treatments'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver