Skip to main content

Advertisement

Table 3 Bias findings from validation studies in InfoUSA and Dun and Bradstreet business lists

From: A step-by-step approach to improve data quality when using commercial business lists to characterize retail food environments

Study Racial/ethnic composition Economic characteristics Urbanicity
InfoUSA Dun and Bradstreet InfoUSA Dun and Bradstreet InfoUSA Dun and Bradstreet
Count accuracy
 Fleischhacker [8] N/A N/A N/A N/A No differences found No differences found
 Liese [15] N/A N/A N/A N/A Urban areas had highest accuracy of stores. Rural areas had lowest accuracy of stores. Suburban areas had the lowest accuracy of restaurants Urban areas had highest accuracy for stores and restaurants. Rural areas had lowest accuracy for stores and restaurants
 Liese [16] Majority white neighborhoods had lowest accuracy No differences found High income and non-poor neighborhoods had lowest accuracy No differences found N/A N/A
 Powell [17] Majority black neighborhoods had lowest accuracy for food stores and restaurants. Majority non-Hispanic neighborhoods has lower accuracy for food stores Majority black neighborhoods had lowest accuracy for restaurants and no difference for food stores No differences found High income areas had lowest accuracy for food stores and no differences for restaurants Urban areas had highest accuracy of stores and restaurants. Rural areas had lowest accuracy of stores and restaurants Urban areas had highest accuracy of stores and restaurants. Rural areas had lowest accuracy of stores and restaurants
Classification accuracy
 Han [23] Majority non-Hispanic and majority black neighborhoods had lowest classification accuracy Majority non-Hispanic and majority black neighborhoods had lowest classification accuracy No differences found No differences found N/A N/A
Locational accuracy
 Liese [15] N/A N/A N/A N/A Urban areas were located with the least distance between observed and listed location. Records in suburban areas were most likely to be allocated to the correct census tract Urban areas were located with the least distance between observed and listed location. Records in suburban areas were most likely to be allocated to the correct census tract