Can scientifically useful hypotheses be tested with correlations?
American Psychologist (2007) - Comment
pubmed: 18020745  doi: 10.1037/0003-066x.62.8.772  issn: 0003-066x  issn: 1935-990x 

Peter M. Bentler