A national manufacturer of ball bearings is experimenting with two different processes, $A$ and $B$ for producing precision ball bearings. It is important that the diameters be as close as possible to an industry standard.
A sample of 9 units from Process A and a sample of 15 units from Process B were randomly taken and their average errors from the industry standard were measured in millimeters. The manager is interested in determining whether there is evidence that the two processes yield different average errors. The population standard deviations are unknown but are assumed to equal.
What should the test statistic be used in testing?