Table 2.

Relative mean squared prediction error based on moments (PE_M) and reconstruction (PE_R) estimates for the three simulation set-ups: case (a) general unsynchronized binary response, case (b) general unsynchronized count response and case (c) unsynchronized binary response with diagonal support analogous to the CDS data. Percentiles of the deviation measures presented are estimated from 200 Monte Carlo runs/data sets.

Case	n	σ2	PEM			PER
			Med	25%	75%	Med	25%	75%
(a)	100	2	.013	.008	.021	.012	.008	.019
(a)	100	4	.013	.008	.022	.012	.007	.019
(a)	100	6	.015	.009	.025	.012	.007	.020
(a)	200	2	.008	.005	.013	.007	.004	.011
(a)	200	4	.009	.006	.015	.007	.004	.011
(a)	200	6	.010	.006	.017	.007	.004	.011
(b)	100	2	.005	.003	.008	.004	.003	.007
(b)	100	4	.005	.003	.009	.004	.002	.006
(b)	100	6	.007	.004	.011	.004	.002	.006
(b)	200	2	.003	.002	.005	.002	.001	.004
(b)	200	4	.004	.002	.006	.002	.002	.004
(b)	200	6	.005	.003	.007	.002	.001	.004
(c)	200	2	.043	.012	.129	.054	.014	.156
(c)	200	4	.050	.013	.161	.055	.015	.160
(c)	200	6	.058	.015	.202	.051	.014	.161
(c)	400	2	.033	.008	.100	.045	.012	.132
(c)	400	4	.042	.011	.137	.047	.013	.138
(c)	400	6	.046	.012	.161	.042	.011	.124