User talk:TeaDrinker/Welcome study

This is very interesting, though I have a comment on the statistical test. Since you are simply comparing proportions in "Number of edits by welcome status" a Fisher exact test is more appropriate, and leads to a finding that the differences are not significant (p=.25). Don't take that as my suggesting welcome messages don't matter (I think they do), but that a more appropriate test does not allow to reject equal chance of editing. TNplinko (talk) 22:03, 26 March 2009 (UTC)[reply]

Interesting, thanks! I don't think the difference in tests is enough to account for the difference (chi-squared is pretty good approximation for large sample sizes). I re-ran the stats using chi-squared and found a similar result. I am going to try and track down what it was I originally did; but I think you found an error here. You're welcome to fix it on the page, if you wish. Thanks and best wishes, --TeaDrinker (talk) 00:06, 29 March 2009 (UTC)[reply]

Curiouser and curiouser; I must have typed a number in wrong somewhere when running the stats; otherwise I have no explanation for the discrepancy. Fixed in the text. --TeaDrinker (talk) 02:54, 17 April 2009 (UTC)[reply]