Some properties of the maximum likelihood strategy for re-pairing a broken random sample (Q1105974)

Matching data from a bivariate population is considered when observations are available only in the form of a broken random sample. In other words, a random sample of n pairs is drawn from the population but the observed data consist of n observations on the second component and the n observations on an unknown permutation of the first component of the n pairs of data. A maximum likelihood matching strategy is revisited. The proportion of approximately correct matches is used to evaluate the perfomance of the pairing strategy as \(n\to \infty\). The small sample behavior of this proportion is studied via a Monte-Carlo simulation in the special case of bivariate normal parent population.