Two criteria for the comparison of multiple arrangements of experiments and their use to answer the question: Is it better to compare p treatment groups with p controls or with a single control
✍ Scribed by Dr. M. Horn
- Publisher
- John Wiley and Sons
- Year
- 1979
- Tongue
- English
- Weight
- 521 KB
- Volume
- 21
- Category
- Article
- ISSN
- 0323-3847
No coin nor oath required. For personal study only.
✦ Synopsis
I n experimental situations examining the effectiveness of p (p z 1) treatmente by comparing treated groups and untreated controls the following question arose: Which experimental arrangement is more appropriate, an arrangement A, using for the p comparisons only one single control, or an arrangement A , using for every comparieon another control. To answer this question two criteria are formulated. The first criterion takes into consideration the expected numbers of type I and type IT errors, and the second criterion the probabilities of type I and of type I1 errom.
According to the first criterion A, and A, do not differ, generally. According to the second criterion A, with the comparison of meam of normal distributed variables is better than A,. Pooling of all p controls is better than A , according to both criteria. A, is better also according to the often used experimentwise type I error rate. The formulated criteria have some analogy to those of MILLER (1966). It may be poaeible to use them not only for comparing experimental arrangements but also tats.