 |
[RC] Statistics but with a bigger sample - better results - Michael MaulStatistics was never my best topics in college. I've learned more about it in the past few days than I wanted to know or relearn.But there's something positive coming out if it. Truman has tripled the sample size for single day rides and can now see MB differences that mean something. In the first analysis - it was assumed that a full season of rides across the whole country would be enough to give a large enough sample for analysis. Truman ran this and said that the variability of the samples in the region was large enough that it was impossible to tell any difference between them. That's what what led to the statement that there were no differences that was posted. But if you continue increasing the sample size - the results should get better. I suggested that we take a much larger sample size of at least 3-6 years. A sample of three years with 27,000 single day ride entries worked. All multiday rides were excluded from the regional LM and MB comparison. RO pulls were looked at and those were relatively uniform across the regions so it doesn't appear that RO would be hiding MB and LM pulls in a region. When that was done - MB differences that were quantifiable were obtained. And these are more in line with what intuition would have told us. NE and SE were the high MB pull regions with SE being 70% larger than the average. NW and MW were about 50% lower than the average. W, PS, CT, MT were near average. The only surprising region for me was CT (slightly below average) where it's hot and humid too. There were no changes in the LM values. A single year was sufficient to establish clear values. The reason for this is that when there are large variations in a small number - which is what happened in the single season MB - you cannot get a consistent value. In the case of LM pulls the variation in the number of pulls was much smaller and a single season was sufficient. The results quoted before still hold. The high regions were SE and NE with up to 50% higher than the average. The only low region was MT with about 50% of the average. So the bottom line here is - take a large enough sample before quoting results. I've certainly relearned that. It was in the stats course I took in college. Truman's whole report is online at http:www.aerc.org/pullstats.pdf It gives you the details and the actual numbers for the pull rates by region and averages. And Trumans's unique description.. :-) He is redoing the the multiday rides on a 3 year basis to compare to the single day rides and that will appear shortly. Mike
Ride Long and Ride Safe!! =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
|