Numerical comparison of maximum credibility estimates with one-step estimates and influence of estimate precision on the distribution of statistics of goodness-of-fit criteria

摘要

Statistical properties of maximum credibility estimates (MLE) and one-step estimates (OS-estimates) obtained as the first iteration of a solution of a set of equations of a probability are explored at final volumes of sampling. It is shown, that at good initial approximations statistical properties of OS-estimations are identical to those of MLE. If the initial approximation appears insufficiently good between distribution laws MLE and OS-estimations there are significations of statistics of goodness-of-fit criteria when verifying complex hypotheses. The estimates computed using the method of moments, usually (but not always) appear good initial approximations for OS-estimations.