Statistics baffle me

April 22, 2005 7:59 AM

Post-hoc stats testing: Is there a difference between the Fisher's PLSD and Fisher's LSD? Or is it just a different name, and the "protected" is implied in the second one? I don't have anybody to ask at work, and my google searching can't explain. And please don't give me advice on a better post hoc test. I know this one is not conservative, but it's not my analysis.
posted by gaspode to Science & Nature (5 answers total)

I don't know much about post-hoc tests, but my guess would be that the 'protected' version incorporates some adjustment of the critical p value, to help avoid detecting differences that are simply due to chance. Can't you just try them both and see if the results differ?
posted by nixxon at 9:50 AM on April 22, 2005

according to this definition they're the same thing.
(i know, you may have googled it, and it's not 100% convincing)
posted by andrew cooke at 9:57 AM on April 22, 2005

sorry, that probably is too cryptic - i have no idea myself, that page is just from googling.
posted by andrew cooke at 9:58 AM on April 22, 2005

Best answer: Any testing procedure that's called 'protected' just refers to the test only being done when the F from the underlying ANOVA is significant. It's generally thought to control error rates, but there's some stats literature that says it doesn't.
posted by milkrate at 11:30 AM on April 22, 2005

I'm also pretty sure they're the same thing.

Bonferroni or bust.
posted by ikkyu2 at 5:55 PM on April 22, 2005