. 2020 Mar 23;15(3):e0229509. doi: 10.1371/journal.pone.0229509

Table 3. Simulation results.

		R²	$σ_{ϵ}^{2}$	$σ_{z}^{2}$	a	c	RMSPE
		R²	$σ_{ϵ}^{2}$	$σ_{z}^{2}$	a	c	LM	SLM	RF	RFRK
1	NL	0.1	9	1	2.94	0.51	3.28 (0.028)	3.22 (0.020)	3.32 (0.027)	3.26 (0.022)
2	NL	0.1	1	9	2.94	0.43	2.77 (0.145)	1.57 (0.029)	2.80 (0.146)	1.64 (0.033)
3	NL	0.9	9	1	2.94	4.56	9.03 (0.074)	9.02 (0.073)	7.45 (0.067)	7.45 (0.065)
4	NL	0.9	1	9	2.94	3.87	7.65 (0.394)	7.40 (0.369)	6.29 (0.322)	5.96 (0.290)
5	L	0.1	9	1	0.33	1.52	3.16 (0.025)	3.09 (0.018)	3.24 (0.026)	3.18 (0.021)
6	L	0.1	1	9	0.33	1.29	2.66 (0.140)	1.34 (0.010)	2.74 (0.143)	1.53 (0.024)
7	L	0.9	9	1	0.33	13.69	4.23 (0.038)	4.19 (0.033)	4.54 (0.036)	4.51 (0.032)
8	L	0.9	1	9	0.33	11.61	3.58 (0.186)	2.82 (0.116)	3.85 (0.201)	3.16 (0.138)

The first column indicates the case number and the second column indicates whether the linear (L) or nonlinear (NL) structural component of the model in (16) dominates. Values for the RMSPE, and parameters a and c, were averaged over 20 independent simulation runs. The standard errors of the RMSPE scores are shown in parenthesis. Note that the standard errors were computed as ${SD}_{RMSPE} / \sqrt{20}$ , where SD_RMSPE is the standard deviation of the RMSPE scores over the 20 runs.