Transcript poster - University of Washington

The SOS Analyst: A GIS Extension for Generating Spatially-Determined Oversampling Rates
Hurvitz, PM1
1University
Background:
Random sampling of individuals or households
generates unbiased estimates of population
parameters. However, obtaining a representative
sample of a minority subpopulation is unlikely
with simple random sampling (e.g., random digit
dialing; see Figure 1). Oversampling is frequently
employed to obtain sufficient variation in
minority populations [1]. Cluster sampling is used
to obtain adequate oversampling rates for
minorities, but requires inflation of sample size.
We introduce a GIS-based method for
determining oversampling rates and probability
weights based on multiple spatially-derived
criteria: the SOS Analyst for ArcView GIS® v
3.2[2].
Figure 6: Oversampling rates attached to point records
of Washington, CAUP <[email protected]>
Steps:
1. Download & install http://gis.washington.edu/phurvitz/sos/download/sos.avx
(Figure 2)
2. Obtain GIS data sets (Figure 3)
1. individual or household points
2. polygons representing spatial oversampling criteria units
3. Define oversampling rates corresponding to desired criteria (Figures 4 & 5)
4. Attach oversampling rates to individual point records (Figure 6)
5. Obtain random sample using oversampling rates (Figure 7)
6. Import point attribute table to statistical software for further analysis
(Figure 8)
4
Oversampling
rates
Figure 7: Random sample selection with oversampling rates
Figure 4: Defining oversampling rates
Figure 2: Download & install SOS Analyst Extension
Figure 1: Racial proportions for
Seattle ZCTAs
Source: US Census 2000
5
3a
1
Sample
frequencies
Population
frequencies
Probability
weights
Figure 3: GIS data sets
Source: US Census 2000, King County GIS
Figure 5: Spatially defined oversampling rates
Figure 7: Importing point attribute table to R
2
Requirements:
ArcView GIS ® v. 3.2 with the Spatial Analyst Extension.
Software and instructions may be downloaded from http://gis.washington.edu/phurvitz/sos/
3b
References:
1. Kalsbeek, W.D., Sampling minority groups in health surveys. Statistics in Medicine, 2003. 22(9): p. 1527-1549.
2. ESRI, ArcView Version 3.2. 1996, ESRI, Inc: Redlands, CA.
6