Table 2.

Mean ($\mu$) and standard deviation ($\sigma$) of the difference between Model A and Model B minus two log likelihood function values over all simulated conditions where the convolution kernel covered 1/2 a cycle of the simulated DLO (i.e., the embedding dimension was 1/2 the number of samples per cycle). ($p<0.05$ indicates the percent of comparisons where this difference exceeded a two degree of freedom chi-square critical value of 5.991 and Cnvrg indicates the number of times both models converged out of 1,000 replications so that the likelihood difference could be calculated)

Simulation					(−2LL MODEL A) – (−2LL MODEL B)
N	SNR	$\eta$	$\zeta$	$\mu$	$\sigma$	$p<0.05$	Cnvrg
50	1	−0.5	−0.1	12.53	6.97	83.9%	937
100	1	−0.5	−0.1	30.85	12.60	99.4%	980
300	1	−0.5	−0.1	154.51	35.75	100.0%	967
50	2	−0.5	−0.1	44.04	14.48	100.0%	937
100	2	−0.5	−0.1	118.31	27.13	100.0%	902
300	2	−0.5	−0.1	599.27	83.53	100.0%	724
50	4	−0.5	−0.1	135.48	25.93	100.0%	418
100	4	−0.5	−0.1	408.41	60.36	100.0%	492
300	4	−0.5	−0.1	155.25	36.48	100.0%	963
50	1	−0.5	−0.3	15.07	7.91	91.6%	953
100	1	−0.5	−0.3	29.77	12.33	98.5%	978
300	1	−0.5	−0.3	95.54	34.85	100.0%	976
50	2	−0.5	−0.3	52.92	15.76	100.0%	929
100	2	−0.5	−0.3	111.76	25.83	100.0%	919
300	2	−0.5	−0.3	195.25	53.13	100.0%	607
50	4	−0.5	−0.3	155.30	26.31	100.0%	245
100	4	−0.5	−0.3	377.83	56.21	100.0%	598
300	4	−0.5	−0.3	1420.39	181.39	100.0%	141