| Message |
further. The series should be double blinded, meaning the persons installing, running and measuring the different ECUs' performance on the dyno should not know which ECU is being tested (probably not possible). Also, even though you are on the right track for "randomization," the power size of your sample is still too small, e.g. more runs needed under a variety of conditions, or better still, under perfectly controlled conditions. That would be if the experiment were perfect, which they rarely are. Design constraints inherent in the problem attempting to be solved often mean a well designed experiment is not feasible or practical. Financial consideration is the other main reason experiments can't get done the way they should. So we are left with the best we can do, but it rarely "proves" the point.
|
 |