On searching probability of two-factor interaction 1

Kashinath Chatterjee, Sudhir Gupta, Angshuman Sarkar
2008 Statistics and Applications   unpublished
The probability of correct model identification, the searching probability, is important for assessing search designs. We show that the expression for searching probability provided by Shirakura, Takashi and Srivastava (1996) significantly overestimates the true searching probability. We first present two classes of balanced designs for identifying one non-negligible two-factor interaction along with their searching probabilities obtained using the expression given by Shirakura, Takashi and
more » ... astava (1996). Then, we present a simulation study which shows that their expression substantially overestimates the true searching probability, particularly for moderate effect size relative to error variance. We, thus, recommend that searching probability should be evaluated through simulation unless error variance is expected to be small relative to effect size. Some results on searching probability for identifying two non-negligible two-factor interactions are also presented.