Quote:
Originally Posted by khafra
A-B-A testing is completely unnecessary. There's various statistical tests that you can apply to your data to find out how credible it is. But if you don't want to do the math, just:
1. Get enough data points in the non-tabbed position that you can fit a bell-curve to them--at least 5, maybe more if they don't cluster nicely at first.
2. Tab up and do the same.
3. Put the two bell curves up on the same graph. The less overlap, the better.
4. If you don't want to do the graphing, just put the raw data points (not overall averages) up here; I or someone else will be happy to do the math.
|
For the math-challenged among us, by "data point" do you mean "#gals consumed for miles driven". Second, your point #2 mentions "tab up". As I stated, those suckers are on to stay and I'm not about to peel them off and put them on again. I can supply 3 tanks of consumption stats prior to installation of tabs. I assume that's what you want?
Ray
