Combining effect size estimates in meta-analysis with repeated measures and independent-groups designs.

In the citing paper

“Hedges g is viewed as more conservative than ES calculations that have previously been recommended for within-participant data.12,28,39”

In the citing paper

“The advantages of using Hedges g for correlated data (cf.39) include the fact that it controls for small sample size bias9 and does not risk inflation of the overall treatment effect when the data are correlated across time in excess of .50.28”

In the citing paper

“We selected Hedges’ g as our ES based on recent guidelines that indicate it is more conservative than previously recommended ES for use in correlated data, particularly for bias associated with small sample size.12,28,39”

In the citing paper

“Effect sizes from pre-post designs are generally less reliable and less informative about the effects of the intervention because they are likely to be influenced by confounding variables (Morris and DeShon, 2002).”

In the citing paper

“Generally, and conform to expectations (Light and Pillemer, 1986; Morris and DeShon, 2002), the average effect of feedback as reported in pre-post design studies is higher than in controlled studies and higher reported in published journal articles than in dissertations.”

In the citing paper

“Effect sizes computed for repeated measures designs and independent groups designs are not directly comparable because different degrees of freedom are used to calculate each [62,63].”

In the citing paper

“When computing effect sizes, we did not correct for the correlation between the pre- and post-treatment means despite the within-subjects design (Morris and DeShon 2002); given that we used the LOCF method to deal with missing data, this would have artificially inflated the pre- to post-treatment mean correlations leading to less conservative estimates of effect size.”

In the citing paper

“The estimation of treatment effects on the dependent variables, for each study, was based on the calculation of the standardized mean change using raw score standardization (SMCR; see Appendix A), taking the standard deviations of the pretest [33,34].”

In the citing paper

“Letting φ() denote the cumulative density function of a standard normal distribution, the interpretation of the standardized mean change can also be facilitated by noting that φ(d) estimates the proportion of individuals for which the difference or change scores reflect a preferable outcome [31].”

In the citing paper

“For an effect size of 0.2 (a “small” effect), the positive interpretation bias is expected to increase (and negative mood is expected to decrease) for φ(0.2) = .58 (i.e., 58%) of individuals.”

In the citing paper

“Effect sizes of 0.5 (a “medium” effect) and 0.8 (a “large” effect) correspond to an increase in positive interpretation bias (and a decrease in negative mood) in 69% and 79% of individuals, respectively [31].”

In the citing paper

“as there was no indication for publication bias [31], we cautiously suggest that even had there been other unpublished data, these would not have changed the results.”

In the citing paper

“Based on the premise for comparing ES previously utilized [58-60], for each of the measures of interest based on the therapeutic intervention (Figure 1).”