Table VII.
Results for example 3
| Univariate (REML) | ρ = 0.7 (REML) | ρ = 0.95 (REML) | Univariate (MM) | ρ = 0.7 (MM) | ρ = 0.95 (MM) | |
|---|---|---|---|---|---|---|
| µ1 | −0.80(0.25) | −0.32 (0.42) | −0.28 (0.31) | −0.80(0.25) | −0.77(0.26) | −0.76 (0.26) |
| [0] | [0.46] | [0.41] | [0] | [0.10] | [0.15] | |
| µ2 | 0.09(0.31) | 0.09(0.31) | 0.10 (0.31) | 0.09 (0.34) | 0.07 (0.34) | 0.06(0.34) |
| [0.64] | [0.63] | [0.62] | [0.70] | [0.71] | [0.71] | |
| κ | 1 | 1 | −1 | −1 | ||
| Max LL | −8.59 | −7.51 | −8.59 | −7.51 |
The parameters µi are the log hazard ratios corresponding to the effects shown in Table III and ρ denotes the common assumed within-study correlation. ‘REML’ indicates that REML has been used and ‘MM’ indicates that the method of moments has been used. Standard errors are in parentheses and the estimated between-study standard deviations which correspond to the parameter in question are shown in square brackets. Max LL denotes the maximum log-likelihood obtained using the within-study correlations shown in a multivariate meta-analysis and κ denotes the estimated between-study correlation.