Talk:Sequential probability ratio test

Theory
Someone might like to expand on my treatment, e.g., to derive the connection between alpha, beta, and the thresholds.--Adoniscik (talk) 19:31, 10 January 2008 (UTC)

I believe the example leaves off (theta0-theta1)/(theta0*theta1) in the final step. 216.164.50.165 (talk) 03:45, 18 November 2009 (UTC)

Theory section: Accept / Reject?
The theory section has these inequalities:


 * $$a < S_i < b$$: continue monitoring (critical inequality)
 * $$S_i \geq b$$: Accept $$H_1$$
 * $$S_i \leq a$$: Accept $$H_0$$

Would it be more correct to say
 * $$S_i \geq b$$: Reject $$H_0$$
 * $$S_i \leq a$$: Accept $$H_0$$

or even
 * $$S_i \geq b$$: Reject $$H_0$$
 * $$S_i \leq a$$: Fail to reject $$H_0$$

The latter would be more in keeping with Hypothesis_testing. — Preceding unsigned comment added by Rajah9 (talk • contribs) 19:45, 3 January 2011 (UTC)

Rajah9 (talk) 14:39, 4 January 2011 (UTC)

Theory section confusing
The use of a and b is contrary to the sense in which Wald uses A and B. Since Wald's paper is authoritative and clear, I see no good reason to deviate from Wald's notation, especially not in a way that might lead to confusion.

In addition, the explanation of the approximation symbols in defining the values of a and b is incorrect or, at least, not the correct explanation. The main reason for the approximation is that these values only approximately guarantee errors of type I and II will occur with probabilities α and β (they only guarantee that at least one will be correct).

MaxSPRT
Is this MaxSPRT algorithm important enough to be included here, or is it just the authors of a paper trying to boost their citations?