A web-based real-estate information and valuation site wants to determine whether it will be worth it to expend time and money to implement a more sophisticated modeling system to estimate home values. A key question is whether the model is more accurate. A firm conducts an experiment in which it tries out the model and obtains valuations on 15 homes that were recently sold (group A). It also samples 15 other homes from its records that were recently sold (group B). It then compares the valuation error in the two groups. To simplify the analysis, it looks first at the absolute value of the valuation error (not caring about sign).
Valuation error (absolute value in $000)
Group A—35, 15, 17, 4, 13, 22, 19, 43, 33, 21, 5, 9, 18, 20, 3.
Group B—23, 44, 9, 34, 39, 49, 22, 22, 7, 11, 29, 17, 35, 6, 11.
Note: To make things easier to present, the sample size shown here, 15 in each group, is smaller than would normally be analyzed by a firm in this situation.