Immediacy and Learning Meta-Analysis 13

instrument, reported effects were adjusted to correct for attenuated measurement (Hunter &

Schmidt, 1990). For 1-item or 2-item measures, an average reliability of .60 was assumed

(Hedges & Olkin, 1985). This adjustment was applied, for example, to effects derived from

perceived learning data gathered through the widely used learning loss measure. After these

corrections were made, both the average effect and the estimate of variance were slightly larger

than originally reported. See Table 1 for the reported or derived correlation (*r*) and the corrected

correlation (*r’*).

______________________

Table 1 about here

______________________

Statistical Analysis

Variance-centered meta-analysis provides a means of objectively comparing average

effects across a group of studies. One advantage of meta-analysis is that the presence of Type II

error can be minimized by computing a weighted average correlation for each study in

accordance with its sample size. Weighting the average effect acknowledges the fact that larger

samples generally provide a more accurate estimate of the true population than do smaller

samples. In addition to average effects, the variance of correlations is also examined for

homogeneity, using a chi-square to compare the observed variance with the expected variance

(Hunter & Schmidt, 1990). A significant chi-square implies the presence of a moderator variable,

which may produce differences between the average effects in subsets of the group of studies.

Results

Comparisons in the overall analysis found a positive and substantial relationship between

overall teacher immediacy and overall student learning (*average r* = .500, *var.* = .037, *k *= 81, *N*