Table 2 Summary of between study variances from the BREM*

From: A case study of an individual participant data meta-analysis of diagnostic accuracy showed that prediction regions represented heterogeneity well

Cutoff date

Parameter

Estimate

Correlation (\({\widehat{\rho }}_{\tau }\))

Area of prediction region

2009

\({\tau }_{0}\)

0.87 (0.53, 1.18)

0.16 (− 0.46, 0.71)

22.83

\({\tau }_{1}\)

0.99 (0.48, 1.42)

2011

\({\tau }_{0}\)

0.87 (0.59, 1.08)

0.32 (− 0.16, 0.71)

16.65

\({\tau }_{1}\)

0.85 (0.54, 1.17)

2013

\({\tau }_{0}\)

0.71 (0.54, 0.88)

0.44 (0.13, 0.73)

13.51

\({\tau }_{1}\)

0.98 (0.69, 1.26)

Full

\({\tau }_{0}\)

0.69 (0.54, 0.82)

0.43 (0.12, 0.71)

11.94

\({\tau }_{1}\)

0.92 (0.67, 1.16)

  1. *All estimates are on the logit scales. Confidence intervals for \({\tau }_{0}\), \({\tau }_{1}, and {\widehat{\rho }}_{\tau }\) were estimated using parametric bootstrap with 1000 replicates.
Back to article page