How to Detect Mistakes in Statistical Analysis

This extremely useful paper reminds us of common statistical mistakes made in articles and papers: ‘Ten common statistical mistakes to watch out for when writing or reviewing a manuscript‘.

Those are:

absence of an adequate control condition or groupinterpreting comparisons between two effects without directly comparing them as a full groupinflating the number of units of analysisspurious correlations (example single weird value)using too small samplescircular analysis (retrospectively selecting features of the data to characterize the dependent variables, resulting in a distortion of the resulting statistical test)too much flexibility of analysisfailure to correct for multiple comparisons in exploratory analysis)over-interpreting non-significant resultsconfusing correlation and causation

Quite a useful checklist to use the next time you review a paper based on statistical analysis!

The post How to Detect Mistakes in Statistical Analysis first appeared on The Fourth Revolution Blog.