Approximate Bayesian computation schemes for parameter inference of discrete stochastic models using simulated likelihood density

Table 1 Comparison of averaged error and mean count number for estimated rate constants over five iterations using algorithms 1 and 2 with simulation number of 10 for system 1.

Δt	*α\k*		1	2	3	4	5
Algorithm 1
3	0.1	MN	15.41	7.21	7.36	8.21	10.05
		AE	0.7668	0.7294	0.7073	0.7832	0.6173
	0.05	MN	175.72	30.66	24.47	28.22	26.5
		AE	0.6120	0.5036	0.5521	0.7175	0.6132
	vary	MN	46.46	25.07	22.76	30.09	88.56
		AE	0.7669	0.5306	0.6780	0.5858	0.5945
5	0.1	MN	26.96	10.47	9.07	11.18	13.19
		AE	0.7107	0.5607	0.5366	0.4693	0.4853
	0.05	MN	130.64	27.38	25.42	35.36	35.79
		AE	0.5826	0.6495	0.4260	0.7548	0.4139
	vary	MN	141.97	30.28	53.47	127.16	2911.58
		AE	0.5587	0.4793	0.5416	0.5960	0.5375
Algorithm 2
3	0.05	MN	467.61	52.34	41.08	69.17	195.69
		AE	0.5834	0.6091	0.4867	0.4995	0.4402
	vary	MN	100.26	32.04	24.78	80.15	1793.64
		AE	0.7132	0.6657	0.6305	0.6705	0.4833
5	0.05	MN	333.17	24.26	32.85	21.11	21.84
		AE	0.5962	0.5340	0.5761	0.4983	0.5518
	vary	MN	243.78	22.6	31.29	34.6	70.25
		AE	0.6565	0.6035	0.5759	0.5488	0.4263

Tests are experimented under different strategies of discrepancy tolerance such as α = 0.1, 0.05 or varies over iterations (AE:Averaged Error; MN: Mean count Number).

ISSN: 1471-2105