Possible Biases in Parameter Estimation with Store Audit Data
研究发现，使用尼尔森双月零售审计数据作为实际销售额的衡量指标，可能导致参数估计偏差高达22%，进而影响研究结论和管理决策。
Researchers frequently use the Nielsen “bimonthly” retail audit data as a measure of actual sales over a two-month period. The results of several simulations suggest that this practice can lead to parameter estimates that are biased by as much as 22%. A bias of this magnitude could seriously affect conclusions reached by researchers and lead managers to make incorrect marketing decisions. This discrepancy is due to the fact that different stores in the Nielsen sample are audited over different two-month periods. As a result, the aggregate Nielsen audit figure partially reflects sales over a three-month period. Several procedures for reducing the bias in the parameter estimates are suggested.