Discussion:
Q: An experimental bumper was designed to reduce damage in low-speed collisions. This bumper was installed on an experimental group of vans in a large fleet, but not on a control group. At the end of a trial period, there were 12 repair incidents (a "repair incident" is an accident that resulted in a repair invoice) for the experimental group and 9 repair incidents for the control group. The dollar cost per repair incident is shown below.
(a) Use MegaStat or MINITAB to perform a two-tailed Mann-Whitney test at α= .05.
(b) Perform a one-tailed parametric t test for two independent sample means by using Excel or MegaStat. Do you get the same decision?
(Data are from Floyd G. Willoughby and Thomas W. Lauer, confidential case study.)
Damage
Old bumper: 1,185, 885, 2,955, 815, 2,852, 1,217, 1,762, 2,592, 1,632
New bumper: 1,973, 403, 509, 2,103, 1,153, 292, 1,916, 1,602, 1,559, 547, 801, 359