摘要：Researchers use Nielsen Homescan data, which provide detailed food-purchase information from a panel of U.S. households, to address a variety of important research topics. However, some question the credibility of the data since the data are self-recorded and the recording process is time-consuming. Matching purchase records from 2004 Homescan data with data obtained from a large grocery retailer, it is evident that quantities purchased are reported more accurately in Homescan than are prices. Many of the price differences may be driven by the way Nielsen imputes prices: when available, Nielsen uses store-level prices instead of the actual price paid by the household. There are also differences by household type in the tendency to make mistakes that are correlated with demographic variables. However, the fraction of variance explained by the documented recording errors is in line with other research data sets for which cross-validation studies have been conducted.
关键词：Nielsen;Homescan;scanner data;validation study
