the value of alpha - this is a rant for sure.

I find myself wrestling with nerdly things... stats until 2 in the morning. No, the quality of my work does not actually improve with reading too many articles about stats, rather I become less and less enamoured of the habits of 'good scientists' who use overly complicated models of strangely transformed parameters to fit the data with a good p value. (sorry that's p value).

Having followed some methods to derive an interesting measure, one famous author publishes in Ecology saying this was above 0.9, and is thus close to 1, whereas 0.85 is not. I dutifully use 8 times the data and ooh, throw the same idea at a t-test - is it different from 1? p says no. But are the other measures that are similar but different also different than 1? p says no. So are the bigger ones (0.88, 0.87, etc) closer to 1 than 0.83? Can't tell. Therefore reviewers will throw me out - this is not significant. So should I thus just report that things are 'closer' or 'farther' from 1? Should I bootstrap, jackknife, simulate and monte carlo my way to significance? Is it okay to say wwwwell my friends, a p of 0.06 is hella less likely than p of 0.3?

Non parametric you say? I saw the sign, and it opened up my mind, I saw the sign... thanks, Ace of Base. Wilcoxon, he says this is much more sensitive. I find the article "the allure of nonparametrics" most pleasing.

Or should I instead befuddle my reviewers and put it all in a likelihood framework? Ah well, yes, friends, let us be brothers, my AIC value is much smaller when this model is used, rather than this one. Do you not know that this is therefore a superior model of what is going on? Even better, shall I disregard any significance and use F-ratios?

And many of my more egg-headed stats-loving friends will say - go Bayesian dude, it changed my world - as though it's some kind of tripped out drug where you achieve knowledge of the prior and thus have universal visions of ecology.

There's a time and a place, my friends, but I'm still a fan of saying hey, this could be a really interesting story. Go out there, collect more data, prove me right or wrong, or find a more sophisticated way to probe this. But wouldn't you like to know?