Techniques for multivariate sample design.

摘要

In this report we consider sampling methods applicable to the multi-product Annual Fuel Oil and Kerosene Sales Report (Form EIA-821) Survey. For years prior to 1989, the purpose of the survey was to produce state-level estimates of total sales volumes for each of five target variables: residential No. 2 distillate, other retail No. 2 distillate, wholesale No. 2 distillate, retail residual, and wholesale residual. For the year 1989, the other retail No. 2 distillate and wholesale No. 2 distillate variables were replaced by a new variable defined to be the maximum of the two. The strata for this variable were crossed with the strata for the residential No. 2 distillate variable, resulting in a single stratified No. 2 distillate variable. Estimation for 1989 focused on the single No. 2 distillate variable and the two residual variables. Sampling accuracy requirements for each product were specified in terms of the coefficients of variation (CVs) for the various estimates based on data taken from recent surveys. The target population for the Form EIA-821 survey includes companies that deliver or sell fuel oil or kerosene to end-users. The Petroleum Product Sales Identification Survey (Form EIA-863) data base and numerous state and commercial lists provide the basis of the sampling frame, which is updated as new data become available. In addition, company/state-level volumes for distillates fuel oil, residual fuel oil, and motor gasoline are added to aid the design and selection process. 30 refs., 50 figs., 10 tabs.