Figure 6.
Circular buffer sizes associated with nonsignficant bias in effect estimates due to point displacement across multiple misclassification levels for ancillary categorical raster data sets. When misclassification rates are less than 20%, use of urban buffer sizes between 1 and 5 km, and rural buffer sizes at or exceeding 5 km resulted in unbiased effect estimates; however when rate of misclassification is greater than 20%, investigators should either refrain from conducting analyses in which data from such surfaces is integrated with DHS data, or should consider more coarse spatial processes when defining covariates of interest.