An investigation in patients with previous myocardial infarction who present with chest pain.

log-likelihood is less than 1.0 are removed from the model. The bootstrap technique is similar to the jackknife technique."4 The basic concept of the technique as applied to the multiple logistic model is as follows. First, the parameters a and A are estimated by maximal likelihood. If the vectors of prognostic variables for the M individuals in the sample used to estimate a and a are respectively x, ..., XM, then a series of random samples of Bennoulli random variables are generated such that Y,j is the ilh random variable in the jth sample Pr[Y1j = 1 ] = F(x,). For each random sample pseudomaximal likelihood estimates of the parameters are obtained. The joint distribution of these pseudomaximal likelihood estimates across the samples can then be used to form confidence limits for a test hypothesis about the parameters. The advantage of this technique over the usual maximal likelihood confidence intervals or the likelihood ratio test is that it does not rely on asymptomatic distribution theory to determine confidence coefficients or significance levels. Asymptotic distribution theory may not be appropriate for small or moderate-sized samples.