08-20-2013, 02:21 PM
(08-20-2013, 12:55 PM)FarmBoy link Wrote:Robertson,
Were you testing whether the difference between 6.3 or 4.5 is statistically significant or whether the difference between 33% (5 of 15 #2 preseason teams ended up ranked #1) and 13% (2 of 15 #1's) is statistically significant?
The test I reported was whether the difference in average finish (6.3 vs. 4.5) was statistically significant. But testing significance for the difference in the probability of finishing #1 (2 of 15 vs. 5 of 15) gives a very similar result (a 39% chance of a difference at least this big happening by chance). Again, not close to being statistically significant. It's really hard to draw solid conclusions based on this small a data set.
[Details for the stats geeks: This p-value comes from Fisher's exact test. If you're one of the folks who believes that test is too conservative, a Pearson Chi-squared gives a p-value of 0.195.]


