Using SAS to Evaluate an Area Sampling Frame for Agricultural Surveys

Sixth International SAS Users Annual Conference

Ron Fecso ; Van Johnson ; and Jeff Geuder ; USDA

PID  :  85097

来源: CEUR

【 摘 要 】

Area frame surveys are a major source of datafor crop and livestock estimates made by theEconomics and Statistics Service (ESS). Frameconstruction errors can result in estimatorbiases, reduce sample efficiency, or both. Toevaluate the quality of the frame, an Area FrameAnalysis Package was developed which uses theStatistical Analysis System~ The analysispackage is intended to be general enough to usefor most ESS frames, and complete enough torequire minimal programming.This paper briefly describes the area framesampling methodology used by ESS and someproblems encountered in frame construction.Examples of some modules in the package areincluded~ along with a brief interpretation ofthe output.BackgroundESS constructs an area sampling frame for eachof the 48 conterminous states (some small statesare combined into one)~ The area frame is an'general purpose f.raMe intended to provide dataon all aspects of agriculturecrop acreage,livestock inventory, agricultural labor, stocksof grain, etc.Using aerial photography and other supplementalmaterial~ the entire land area in a state isgrouped into several strata. The stratificationvariable for the frame is the extent of cultiva tion (0100%). The number of strata varies bystate. depending on the state's land use andgeography. Each land use stratum is thendivided into primary sampling units (PSU's).The PSU's are grouped into what ESS has termedflpaper strata, which contain, to the extentpossible~ adjacent areas of land with similaragriculture. A replicated sample of these PSUfsis then selected with probability proportionalto size. Within each selected PSU~ the finalstage sampling units, or segments are delin eated. An equal number of segments is selectedwithin each paper stratum within the land usestratum. The analysis package can be used todetermine the optimal number of paper strataand replications.Obstacles Encountered in Frame ConstructionThe step from theory and design to applicationin area frame surveys is cpmplicated by severalobstacles. Among these are variability insegment size, the need for uptodate mappingmaterials, and the subjective interpretationof aerial photography.Although optimum segment size can bedetermined statistically. area frame segmentscannot always be defined to meet the optimum.Segment boundaries must be physicalcharacteristics easily identifiable to an363enumerator during data collection. In manyparts of the' country, segments with an optimumsize of 1 square mile may have to be drawn as.5 square mile or l~5 square miles simplybecause no other segment boundaries areavailable. The resulting variat~on in segmentsize can considerably reduce efficiency andadd bias.Frame construction is also affected by outof date maps and photographs~ Constantly changingland usage makes photography obsolete in a veryshort period of time. Misstratification occurswhen a