ESTIMATING THE SIZE OF POPULATION FROM A SINGLE SAMPLE

点击翻译按钮获取中文摘要

Estimating the size of a population of individuals usually requires either a full census or samples on multiple occasions - a capture- recapture census. However, for many populations a full census may be impossible, and the two-stage approach may be unsatisfactory. Marking may not be possible, the cost of mounting separate capture efforts may be large, matching individuals in the two samples may be difficult, or the assumption of closure (no additions and no deletions between captures) may be questionable. Laska, Meisner and Siegel (1989) and Laska and Meisner (1992) have developed two methods, that require only a single sample, that overcome many of the problems. In the first approach, it is desired to estimate the distinct number of persons who engage in a specific activity during a specified period of time, e.g., individuals who received treatment in any mental health facility in a state during 1992. During a survey, information is obtained as to when individuals were last associated with the activity, e.g., for patients who came for service during the survey week, the number of weeks since they last came for service is ascertained. In the second approach called the plant-capture method, it is desired to estimate the size of a population at a single point in time. A known number of individuals are planted, and the (usual) recapture phase is performed. Both of these methods enable estimation of quantity, p, analogous to the probability of capture, which, in turn, enables estimation of the population size. We propose to extend and refine single capture methods. This includes statistical approaches to handle multiple sites, subpopulations, tests of bias, reducing the sample size and combining the two single sample methods. Through consultations these will be specifically applied to three target mental health populations: (i) the number of homeless and the mentally ill homeless in an area; (ii) the unduplicated number of patients receiving service in a year in a mental health jurisdiction; and (iii) the proportion of patients utilizing a general health care facility who have psychiatric disorders. Single sample methods may be applicable to estimating the size of many other mental health target groups such as, e.g., the chronically mentally ill who reside in the community, and mentally ill chemical abusers.