On estimating the parameter of a truncated geometric distribution by the method of moments

Abstract : The purpose of this note is to present a moment estimator for the parameter of the geometric distribution from samples which are truncated at arbitrary points in either or both tails of the distribution. This is derived by a modified method of moments procedure suggested for use with truncated distributions by Rider in his 1953 and 1955 papers in JASA. Its efficiency is compared relative to the maximum likelihood estimator given by Thomasson and Kapadia in Annals of the Institute of Statistical Mathematics, 1968. The advantage of this method, as compared to maximum likelihood is that it yields an explicit solution for the estimator, so that the computation of the estimator does not require the solution of a high-degree polynomial. This estimator should be particularly useful when Table I given by Thomasson and Kapadia cannot be used; for example when greater accuracy is required or when d > 27. (Author)