Correcting for Unreliability in Test Measures

So far this meta-analysis technique has corrected for one source of error, sampling error. There are two other forms of error. The other two are measurement error, and range restriction.

Measurement Error or Test reliability is assessed by using the two reliabilities which apply to the study: r_xx and r_yy. While you can correct the individual study statistics for range variation, all too often this information is not available at the study level. Hunter and Schmidt (Hunter et. al, 1982 and Hunter and Schmidt, 1990) present a method of correcting the sample weighted mean effect size using a distribution of reliability estimates. A distinct advantage to this method is that it does not depend on using jus the studies directly involved in the meta-analysis. Any study that assesses the pertinant reliability estimates can be used to construct the reliability distribution.