Sunday, March 17, 2013

The Significance of Statistical Significance

The other year I authored a paper that takes unique approaches to demonstrating that the commonly used term for statistical significance (p < .05) is obtainable from data that is not 'significant'. The paper was invited for publication in a new journal, but at the time I was too busy focusing on other endeavours to develop the paper and take advantage of the publication invitation.

 
I have been kicking myself for over a year for not taking advantage of the invitation. Fortunately my self punishment on this issue is at an end as another journal has invited the paper for publication. The paper has been published and you can access it here. The paper is meant to educate the general scientist by taking unique approaches to demonstrating that statistical significance can be obtained from data that is qualitatively insignificant. Without a reasonable understanding of this issue, scientists who are overconfidant in the results of their statistical analyses can erroneously conclude that their own work is more significant than it truely is.

This paper was authored very much for the general scientist who has only a cursory interest in statistics, specifically those interested in how statistics should or should not influence our conclusions regarding the 'significance' of one's own experiments. It was not intended for statistician types as the paper avoids using terms such as the Null Hypothesis and instead uses the most accessible language available, in order to potentially reach as large an audience as possible.