CIAAWconsensus (version 1.3)

mmm: Multivariate meta-analysis of correlated effects

Description

This function provides meta-analysis of multivariate correlated data using the marginal method of moments with working independence assumption as described by Chen et al (2016). As such, the meta-analysis does not require correlations between the outcomes within each dataset.

Usage

mmm(y, uy, knha = TRUE, verbose = TRUE)

Arguments

y

A matrix of results from each of the n laboratories (rows) where each study reports m isotope ratios (columns)

uy

A matrix with uncertainties of the results given in y

knha

(Logical) Allows for the adjustment of consensus uncertainties using the Birge ratio (Knapp-Hartung adjustment)

verbose

(Logical) Requests annotated summary output of the results

Value

studies

The number of independent studies

beta

The consensus estimates for all outcomes

beta.u

Standard uncertainties of the consensus estimates

beta.U95

Expanded uncertainties of the consensus estimates corresponding to 95% confidence

beta.cov

Covariance matrix of the consensus estimates

beta.cor

Correlation matrix of the consensus estimates

H

Birge ratios (Knapp-Hartung adjustment) which were applied to adjust the standard uncertainties of each consensus outcome

I2

Relative total variability due to heterogeneity (in percent) for each outcome

Details

The marginal method of moments delivers the inference for correlated effect sizes using multiple univariate meta-analyses.

References

Y. Chen, Y. Cai, C. Hong, and D. Jackson (2016) Inference for correlated effect sizes using multiple univariate meta-analyses. Statistics in Medicine, 35, 1405-1422

J. Meija and A. Possolo (2017) Data reduction framework for standard atomic weights and isotopic compositions of the elements. Metrologia, 54, 229-238

Examples

# NOT RUN {
## Consensus isotope amount ratios for platinum
df=normalize.ratios(platinum.data, "platinum", "195Pt")
mmm(df$R, df$u.R)
# }