I'd also like to point out some reasons we should be skeptical of the results in the above test. (Note I'm not saying "dismiss" his results, but why we should still question them.)
Also: the guy who posted the test deserves a lot of credit for thoroughly describing what he did.
1) The tester did not "remove himself" from the experiment, exposing the results to the potential effects of
experimenter bias.
Experimenter bias risk could have been avoided by recording speed digitally, which also would have been more accurate, since the sampling rate of the recording could have then been used to time the runs (see below).
The potential for bias exists at 4 points: in the "reading/interpreting" of the analog speedometer; depressing the clutch & starting the timer; reading the speedometer at the end of the run; stopping the timer.
2) The accuracy of the results is questionable due to reaction time of the experimenter on the timer. Average human reaction time on a similar "clicking" task is about 0.22 seconds. (eg.
Human Benchmark - Reaction Time Stats ).
Consider the average difference of the runs was 0.61 seconds, and you can see why this is important.
The experimenter himself points out the difficulty of being accurate:
Quote:
|
starting the timer exactly simultaneously with pushing the clutch pedal took me a while to learn.
|
3) The test is complicated by placing additional VG's on the side of the car for the second "B" set of runs. It's not really an A-B-A-B test, it's an A-B-A-C test.
If you compare the numbers as 2 separate tests, you may get different results. EG: why did doubling the number of VG's apparently
decrease coasting time between the two B runs. (The experimenter suggests it's due to light rain at the end of the second test.)
4) The experiment was also run in the opposite direction, and the results turned out to be not statistically significant:
Quote:
|
on the way back from each run, I measured the time in the other direction. In this other direction, whatever difference the airtabs made was not statistically significant, though [the raw data] still showed a slight gain.
|