TABLE 4.
Performance of our composite-likelihood approach at estimating σ for the spatial model
| β = 1
|
β = 2
|
||||||
|---|---|---|---|---|---|---|---|
| m | θ |  |
RMSE | Coverage (%) |  |
RMSE | Coverage (%) |
| 20 | 15 | 1.01 | 0.22 | 96 | 1.02 | 0.25 | 93 |
| 20 | 30 | 0.98 | 0.19 | 95 | 1.00 | 0.20 | 95 |
| 40 | 15 | 0.94 | 0.23 | 92 | 0.99 | 0.24 | 95 |
| 40 | 30 | 0.96 | 0.21 | 91 | 0.93 | 0.18 | 95 |
We report the mean of the estimates of σ (truth σ = 1), the root mean square error of the estimates, and the coverage probability of 95% approximate confidence intervals. (We split the data into disjoint subsamples of size 10. The grid of σ-values ranged from 0 to 2.)