hourlens
hourlens
Jul 27, 2017 · 1 min read

My take from your perspective is that clinical significance requires statistical significance, and that it is actually less likely for something to be clinically significant as opposed to statistically significant. So clinical significance is of “higher order” than statistical significance.

However, I hypothesize that there is clinical significance without statistical significance. At least for me, that’s what the 10 years of experience / education you mentioned do with an expert — enabling pattern recognition, and some patterns have clinical significance without statistical significance. What do you think about that?

Something that would oppose this argument is the fact that today’s best pattern recognition entities — machine learning algorithms — base their abilities on statistical significance and advanced probability calculations.

So in conclusion, with complete information in a system, statistical and clinical significance should converge and eventually be identical, right? The difference between clinical significance and statistical significance then is a lack of information, i.e. variables not taken into account even in multivariate analyses.

    hourlens

    Written by

    hourlens