The several principal experimental matrices of the Auto/Oil Air Quality Improvement Research Program (AQIRP) were statistically designed as regards vehicle fleet size and fuel property combinations. The test results were analyzed using powerful standard statistical methods to extract the maximum amount of information from the data. The analysis included the use of appropriate data transformations and also graphical methods to display the results.The test fleets were sized to control the risk of failing to detect an important effect while providing assurance that unimportant, small effects have little chance to be found highly significant. To do this required estimation of the pertinent error term from other test programs which did not completely correspond to the type of testing contemplated in the AQIRP. These estimates turned out to be quite good.The compositional fuel matrices were blocked into two fuel groups. This randomized blocked design allowed preliminary estimation of the average fuel property effects to be made at the conclusion of Block 1 testing. In addition, blocking into two fuel groups, and the use of an “Industry Average” fuel before, during, between, and after the two randomized blocks, improved the precision of the observed fuel property effects.Running a pair of repeat tests for each fuel-vehicle combination provided a quality control check on the vehicles and the entire FTP measurement system. A third test was run when the difference of the first two results exceeded a predetermined limit, permitting the use of an outlier detection procedure to identify those tests that were significantly out of line.The use of standard experimental designs also meant that the analysis and interpretation of the results were reasonably straightforward.