Resumen de Models of not-so-good behavior: Yet another way to squeeze causality and recommendations for practice out of correlational data
Alyssa L. Reinhart, Samuel H. Haring, Joel R. Levin, Erika A. Patall, Daniel H. Robinson
Two previous studies examining 5 empirical educational psychology research journals (Hsieh et al., 2005; Robinson, Levin, Thomas, Pituch, & Vaughn, 2007) found that in the 21-year period from 1983 to 2004, there was a decrease in intervention and randomized experimental research, whereas in the 10-year period from 1994 to 2004, there was an increase in recommendations for practice based on nonintervention observational/correlational research. The present study extends this research to determine the extent to which statistical modeling analyses were conducted in articles published in the 10-year period from 2000 to 2010. The aforementioned trends continued: In 2010, only 23% of the published empirical research studies used random assignment, and 46% of nonintervention observational/correlational articles included recommendations for practice. Additionally, the percentage of the latter articles that used statistical modeling analyses increased from 15% in 2000 to 54% in 2010. Moreover, across the 2 time periods, observational/correlational articles that incorporated modeling analyses were about 1.7 times more likely to contain recommendations for practice than such articles that did not incorporate modeling analyses (59% vs. 35%, respectively). These findings suggest that educational researchers may be overstepping the warrants of statistical modeling techniques by using them to confirm, rather than simply to disconfirm, causal hypotheses derived from correlational data (Whitehurst, 2003).
