EFFECT OF SAMPLE SIZE ON DISAGGREGATE CHOICE MODEL ESTIMATION AND PREDICTION

摘要: Sampling error is one of several types in econometric modeling. The relationship between sampling and sample size well known for both estimation prediction. objective this paper to provide an empirical foundation using these relationships guide researchers planners the determination model development. Analytic are formulated size, precision parameter estimates, replication parent population, alternative (transfer) population. Application case indicates that sizes required obtain reasonably precise estimates substantially larger than generally considered be needed disaggregate estimation. Nevertheless, appear adequate obtaining accurate observed choice behavior corresponding results prediction a different population complicated by issue intrapopulation transferability. Although reported should validated other contexts, it appears requires use samples formerly believed. Samples on order 1,000 2,000 observations may relatively simple models. some reduction requirement obtained improved design, unlikely final requirements can reduced less observations.