A Sequential Statistical Analysis for Word Articulation Tests
Mones E. Hawley, Walter Wong, Willard F. Meeker · The Journal of the Acoustical Society of America · 1953
The problem of deciding when the intelligibility of one speech communication system is equivalent to the intelligibility of another system has been solved utilizing the techniques of sequential statistical analysis. If the experimenter is willing to tolerate some arbitrary probability (a) that the method will reject the system when it should be accepted and some arbitrary probability (b) that the method will accept the system when it should be rejected, he may make a rapid, running analysis which will indicate whether he should accept the system, reject it, or continue testing. The system should be accepted when σ22 ∑ α=1m x1α − σ12 ∑ α=1m x2α≦σ12σ22 loge[b/(1 − a)] + m(σ22 − σ12) (μ1 − μ2)/2. The system should be rejected when σ22 ∑ α=1m x1α − σ12 ∑ α=1m x2α > σ12σ22 log[(1 − b)/a] + m(σ22 − σ12) (μ1 + μ2)/2. where, in both cases, σ1 = standard deviation of the scores of the first system, σ2 = standard deviation of the scores of the second system, m = number of the tests, x1α = score of the first system on the αth test, x2α = score of the second system on the αth test, μ1 = mean value of the scores of the first system, and μ2 = mean value of the scores of the second system. If neither condition is satisfied, testing should continue. The method checks with the longer analysis of variance which is used at present.