Skip to main content

Table 3 Examples of dependence between two sampling variances (v1 and v2) and their covariance for four common effect size statistics

From: Quantitative evidence synthesis: a practical guide on meta-analysis, meta-regression, and publication bias tests for environmental sciences

Effect size

Situation

Variances estimate

Covariance estimate

Proportion

Shared measurement

\({v}_{1}=\frac{{y}_{1}\left({n}_{1}-{y}_{1}\right)}{{n}_{1}^{3}}\)

\({v}_{2}=\frac{{y}_{2}\left({n}_{2}-{y}_{2}\right)}{{n}_{2}^{3}}\)

\(\rho \sqrt{\frac{{y}_{1}\left({n}_{1}-{y}_{1}\right)}{{n}_{1}^{3}}\frac{{y}_{2}\left({n}_{2}-{y}_{2}\right)}{{n}_{2}^{3}}}\)

Zr

Shared measurement

\({v}_{1}=\frac{1}{2}{\text{ln}}\left(\frac{1+{r}_{1}}{1-{r}_{1}}\right)\)

\({v}_{2}=\frac{1}{2}{\text{ln}}\left(\frac{1+{r}_{2}}{1-{r}_{2}}\right)\)

\(\rho \sqrt{\frac{1}{4}{\text{ln}}\left(\frac{1+{r}_{1}}{1-{r}_{1}}\right){\text{ln}}\left(\frac{1+{r}_{2}}{1-{r}_{2}}\right)}\)

lnRR

Shared measurement

\(v_{1} = \frac{{s_{1C}^{2} }}{{n_{1C} \overline{x}_{1C}^{2} }} + \frac{{s_{1T}^{2} }}{{n_{1T} \overline{x}_{1T}^{2} }}\)

\(v_{2} = \frac{{s_{2C}^{2} }}{{n_{2C} \cdot \overline{x}_{2C}^{2} }} + \frac{{s_{2T}^{2} }}{{n_{2T} \cdot \overline{x}_{2T}^{2} }}\)

\(\rho \sqrt {\left( {\frac{{s_{1C}^{2} }}{{n_{1C} \overline{x}_{1C}^{2} }} + \frac{{s_{1T}^{2} }}{{n_{1T} \overline{x}_{1T}^{2} }}} \right)\left( {\frac{{s_{2C}^{2} }}{{n_{2C} \overline{x}_{2C}^{2} }} + \frac{{s_{2T}^{2} }}{{n_{2T} \overline{x}_{2T}^{2} }}} \right)}\)

 

Shared control

\(v_{1} = \frac{{s_{1C}^{2} }}{{n_{1C} \overline{x}_{1C}^{2} }} + \frac{{s_{1T}^{2} }}{{n_{1T} \overline{x}_{1T}^{2} }}\)

\(v_{2} = \frac{{s_{1C}^{2} }}{{n_{1C} \overline{x}_{1C}^{2} }} + \frac{{s_{2T}^{2} }}{{n_{2T} \overline{x}_{2T}^{2} }}\)

\(\frac{{s_{1C}^{2} }}{{n_{1C} \overline{x}_{1C}^{2} }}\)

SMD

Shared measurement

\({v}_{1}=\frac{1}{{n}_{1C}}+\frac{1}{{n}_{1T}}+\frac{{d}_{1}^{2}}{2\left({n}_{1C}+{n}_{1T}\right)}\)

\({v}_{2}=\frac{1}{{n}_{2C}}+\frac{1}{{n}_{2T}}+\frac{{d}_{1}^{2}}{2\left({n}_{2C}+{n}_{2T}\right)}\)

\(\rho \sqrt{\left(\frac{1}{{n}_{1C}}+\frac{1}{{n}_{1T}}+\frac{{d}_{1}^{2}}{2\left({n}_{1C}+{n}_{1T}\right)}\right)\left(\frac{1}{{n}_{2C}}+\frac{1}{{n}_{2T}}+\frac{{d}_{2}^{2}}{2\left({n}_{2C}+{n}_{2T}\right)}\right)}\)

 

Shared control

\({v}_{1}=\frac{1}{{n}_{1C}}+\frac{1}{{n}_{1T}}+\frac{{d}_{1}^{2}}{2\left({n}_{1C}+{n}_{1T}+{n}_{2T}\right)}\)

\({v}_{2}=\frac{1}{{n}_{1C}}+\frac{1}{{n}_{2T}}+\frac{{d}_{2}^{2}}{2\left({n}_{1C}+{n}_{1T}+{n}_{2T}\right)}\)

\(\frac{1}{{n}_{1C}}+\frac{{d}_{1}{d}_{2}}{2\left({n}_{1C}+{n}_{1T}+{n}_{2T}\right)}\)

  1. For the 2nd column, see Fig. 2. For the 3rd and 4th column, notations represent: the subscript 1C and 2C (control group for 1st and 2nd effect size, respectively, but for shared control, 1C is used for both effect sizes, but 1C and 2C are the same cohort or set of plots), the subscript 1T and 2T (treatment group for the 1st and 2nd effect size, respectively; for shared groups, 1T and 2T represents different groups of individuals/plots whereas, for shared measurements, 1T and 2T are the same set of individuals/plots), \(\rho \) is a correlation in sampling error variance between two measurements, and the other notations are as in Table 1 and the main text (see also [54, 55])