NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Johnson BT, Huedo-Medina TB. Meta-Analytic Statistical Inferences for Continuous Measure Outcomes as a Function of Effect Size Metric and Other Assumptions [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2013 Apr.

Cover of Meta-Analytic Statistical Inferences for Continuous Measure Outcomes as a Function of Effect Size Metric and Other Assumptions

Meta-Analytic Statistical Inferences for Continuous Measure Outcomes as a Function of Effect Size Metric and Other Assumptions [Internet].

Show details

Glossary of Terms

Bias

The extent to which the observed UMD or SMD differs from the parametric value. Positive values of bias imply over-estimations of the parametric effect size and negative values imply under-estimations.

Change-score metric

The difference between two repeated measures compared with the variability of change scores.

Coverage

The proportion of replications for which the 95% confidence interval for each index did not include the null value, ES = SMD = UMD = 0.

Effect size (ES)

The magnitude or degree of the association between two variables. In the current investigation, comparisons between groups, across time, or both, are used (e.g., standardized mean difference; unstandardized mean difference).

Efficiency

A measure of the optimality of an estimator that reaches the closest value to the parameter with the minimum variance. To the extent that efficiency is positive, statistical power is maximized to detect the parametric value.

Mean square error (MSE)

A measure of the average of the square of the errors that evaluates the quality of an estimator in terms of its variation and unbiasedness. Bias and efficiency are, in effect, components of the MSE.

One-group repeated-measures design

A study methodology in which a single sample is observed at two or more time points (e.g., before and after a treatment).

Pooled standard deviation

The sample-size weighted mean standard deviations of two or more groups.

Raw-score metric

A metric that compares a mean or mean difference between conditions or times with the variability of scores within each condition.

Sampling variance of effect size for one-group repeated-measures design

The variance of the sampling distribution, which is the distribution of values that result from repeated random samples of the same size using a repeated-measures design (see Table 3 for extant estimations of this statistic).

Sampling variance of effect size for two independent groups

The same as the preceding one but for studies following a two-groups design, so using two independent samples (see Table 4 for extant estimations of this statistic).

Standardized mean difference (SMD) effect size

The difference of two means divided by the pooled standard deviation.

Standardized mean difference (SMD) for one-group repeated-measures design

The effect size comparing two means at different times for the same group relative to the standard deviation (see Table 1 for extant estimations of this statistic).

Standardized mean difference (SMD) for two independent groups

The effect size reflecting the change in means for two independent groups (repeated-measures, between-groups version) or the comparison of the means at post-test for two independent groups (between-groups version). (See Table 2 for extant estimations of this statistic.)

Two-groups repeated measures design

A study methodology in which two groups (or arms; e.g., treatment and control) are observed at two or more times.

Unstandardized mean difference (UMD) effect size

The difference of the two means in their original metric or scale.

Following are the Greek terms that appear in this report

TermDefinition
δThe parametric difference between two groups
δ̂jThe sample estimate of population parameter δ for the jth replication
μpostEParametric mean at post-test for the experimental group
μPreEParametric mean at pretest for the experimental group
μPostCParametric mean at post-test for the control group
μPreCParametric mean at pretest for the control group
σPre2Parametric variance at pretest
σPost2Parametric variance at post-test
σPre,PostCovariance between pre- and post-test
ρParametric correlation
ρpre-postParametric correlation between pre-and post-test
σCParametric standard deviation of the control group
σEParametric standard deviation of the experimental group
τ2Between-study variance

Following are the Latin abbreviations used in this report

TermDefinition
dbStandardized mean difference proposed by Becker (See Table 1, No. 3 for details and elements of the equation)
db_nonrBecker’s standardized mean difference, excluding the correlation factor 2(1 – r) in its variance estimation
dfDegrees of freedom
dgStandardized mean difference proposed by Gibbons (See Table 1, No. 4 for details and elements of the equation)
dhbStandardized mean difference proposed by Hedges (See Table 2, No. 6 for details and elements of the equation)
dhwStandardized mean difference proposed by Huedo-Medina & Johnson (See Table 1, No. 5 for details and elements of the equation)
dhw_nonrHedges’ standardized mean difference using the within-study degrees of freedom and excluding the correlation factor 2(1 – r) in its variance estimation.
ds1Standardized mean difference proposed by Shadish (See Table 2, No. 10 for details and elements of the equation)
ds2Standardized mean difference proposed by Shadish using the standard deviation from ANOVA results (See Table 2, No. 11 for details and elements of the equation)
ds3Standardized mean difference proposed by Shadish using the standard deviation from ANCOVA results (See Table 2, No. 12 for details and elements of the equation)
dtchStandardized mean difference based on t-test for change-score metric (See Table 1, No. 2 for details and elements of the equation)
dtraStandardized mean difference based on t-test for raw-score metric (See Table 1, No. 1 for details and elements of the equation)
ESEffect size
HAM-DHamilton rating scale of depression
kNumber of studies
M HAM-DMean score on the HAM-D
MCMean for control group
MEMean for experimental group
mmHgMillimeters of mercury (used in measures of blood pressure)
MSEMean square error
NTotal sample size
nGroup sample size
OROdds ratio
rEstimated correlation
RnsThe number of replications
SDStandard Deviation
SEUMDThe standard error for the UMD
SESMDThe standard error for the SMD
SMDStandardized mean difference (d)
UMDUnstandardized mean difference (Equation 19, Table 5)
varone-g (db)Variance estimate for one group design with repeated measures of the standardized mean difference proposed by Becker (See Table 3, No. 13 for details and elements of the equation)
varone-g (dg)Variance estimate for one group design with repeated measures of the standardized mean difference proposed by Gibbons (See Table 3, No. 14 for details and elements of the equation)
vartwo-g (db)Variance estimate for two group design with repeated measures of the standardized mean difference as a function of two effect sizes proposed by Becker (See Table 4, No. 16 for details and elements of the equation)
vartwo-g (db_t)Variance estimate for two group design with repeated measures of the standardized mean difference proposed by Hedges (See Table 4, No. 15 for details and elements of the equation)
vartwo-g (dg)Variance estimate for two group design with repeated measures of the standardized mean difference as a function of two effect sizes proposed by Gibbons (See Table 4, No. 18 for details and elements of the equation)
vartwo-g (dg_t)Variance estimate for two group design with repeated measures of the standardized mean difference proposed by Gibbons (See Table 4, No. 17 for details and elements of the equation)
VarUMDThe variance estimate for the UMD
VarSMDThe variance estimate for the SMD
YCControl outcome
YEExperimental outcome

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...