St. Pat's day, revisited

Abstract: A statistical test for differences in proportions indicates that people at my university campus do indeed wear more green on St. Pat's day than on other days.


Back on St. Patrick's Day, I wrote a blog entry where I wondered whether sufficient Aussies (or at least, Brisbanites) acknowledged St. Pat's Day. On that day I did a rough sampling of how many people I saw who were wearing green. I counted 23 out of 151 -- which is 15.2%.

Today, I thought I'd do another quick tally -- even though I **should've** done it on another Friday, not a Tuesday [today] -- but, eh! (Can't be bothered. So much for scientific rigor!)

Today the count was 14 out of 164, or 8.5%

Using the ''immediate form of two-sample test of proportion'' (the prtesti command) in Stata (a CLI-based stats package), it turns out that the probability of the difference between 8.5% and 15.2% being due to a ''real'' underlying difference -- as opposed to just a fluke of grabbing a strange sample -- was 6.4% (i.e. p=0.064).

That means that there's a 93.6% chance of people wearing more green on St. Pat's day than on a ''normal'' day is due to a systematic difference between the days -- like, it being St. Pat's Day!!!

In fact, the results for a one-tailed test -- that is, my assuming that St. Pat's day would have more green, rather than less green -- was 3.2% (p=0.032), or a 96.8% of being a ''solid, believable'' result.

Right on!!!

They do believe!!! :)


--GG