Horseshoes, hand grenades, and pregnancy

More from Rocket Science (it's a good week).  Suppose you've put a lot of time and effort into a research project, and you've chosen for your own reasons to use the arbitrary but time-honored and respectable standard of statistical significance (i.e., p = 0.05, but your results stubbornly refuse to make the grade.  You could instead quote the effect size with a confidence interval, but for whatever reason you refuse to do that.  All that time and effort, with nothing to publish, seems unfair.

Not to worry.  New, substance-free circumlocutions have sprung up to describe how tantalizingly close your work has come to rendering meaningful results.  Psychologically Flawed has published a convenient list in alphabetical order, including such evasive or emotionally charged winners as:
"flirting with conventional levels of significance (p=0.1)"
"inconclusively significant (p=0.070)" 
"narrowly missing conventional significance (p=0.054)" 
"nearly borderline significance (p=0.052)" 
"not absolutely significant but very probably so (p=0.05)" 
"only slightly missed the conventional threshold of significance (p=0.062)" 
"teetering on the brink of significance (p=0.06)" 
"trend bordering on statistical significance (p=0.066)" 
"very closely brushed the limit of statistical significance (p=0.051)"
They coulda been a contenda!  My favorite: "not significantly significant but . . . clinically meaningful (p=0.072)."  I look forward to papers describing results as "longing for significance but thwarted by hidebound, linear, and cruelly normative conventional standards."

1 comment:

E Hines said...

Not to be too pedantic, but that's not p = 0.05, but p < 0.05. There. Finally got to put my minor in statistics to actual, meaningful use.

The last phrase, about clinically useful, with a proper discussion might actually be legitimate rather than obfuscatory, if there's discussion of the meaning of that phrase, of the arbitrary nature of "significance," and of the results themselves.

But then, most publication efforts are about being seen and read for tenure and other personal gain, rather than disseminating actual knowledge. Thus the need to weasel-word outcomes, so as to keep the funding flowing.

Eric Hines