SELECTING THE BEST ONE OF SEVERAL BINOMIAL POPULATIONS 



563 



required number of observations. In the example of the preceding sec- 

 tion the value of n required for the alternative specification is 45 as 

 compared to 51 observations per process required to satisfy the original 

 specification with the same P* and with d* = p*i) — 'p*2]. Here the 

 saving is only moderate. The saving will be much larger if pfij and 



1. 00 



0.95 



0.90 



0.85 



0.80 



0.75 



0.70 



0.65 



0.60 



0.55 



0.50 



0.45 



0.40 



0.35 



0.30 



0.25 



0.05 0.10 0.15 0.20 025 030 0.35 Q40 0.45 0.50 



d 



Fig. 20 — Probability of a correct selection as a function of the true difference 

 f^ = P[i) — p[2] under the least favorable configuration for A: = 4 and selected 

 values of n. 



