Table 1. Summaries of statistical tests comparing models of aphid motion using order parameters.

Exp	P		Mang		Mabs		da		Mov%
	D	R95%	D	R95%	D	R95%	D	R95%	D	R95%
1	0.17	0.13	1.50	0.16	-2.00	0.15	0.26	0.05	20.2	0.40
2	-0.25	0.10	-0.97	0.11	-1.66	0.09	0.25	0.02	0.78	0.24
3	1.09	0.10	0.40	0.11	-1.71	0.10	0.54	0.04	25.1	0.26
4	1.70	0.08	-0.78	0.09	-1.24	0.07	0.16	0.02	1.68	0.21
5	-2.22	0.09	-0.63	0.09	-1.10	0.07	0.31	0.03	7.68	0.24
6	1.04	0.09	-0.49	0.10	-1.68	0.09	0.53	0.04	24.0	0.26
7	-0.97	0.11	-1.28	0.13	-1.99	0.11	0.57	0.04	23.9	0.29
8	-3.30	0.25	-2.11	0.22	-2.87	0.26	0.57	0.25	13.2	0.63
9	-4.69	0.25	-3.14	0.22	-2.81	0.20	0.48	0.15	11.7	0.34

Data to the left of the thick vertical line summarize statistical tests performed on three order parameters that do not use a priori knowledge of the models: polarization P, angular momentum M_ang, and absolute angular momentum M_abs. Data to the right of the thick vertical line correspond to two order parameters that do use a priori knowledge of the models: average distance to nearest neighbor d_a and overall percentage of insects moving Mov_%. For each order parameter and each experiment, we calculate the average Euclidean distance D_int between the experimental data and data produced by 100 runs of the interactive model seeded with the same initial condition as the experiment. We also calculate D_con, which is similar, but for the control model. See Eqs (10) and (11) and main text for more detail. To assess whether one model produces data closer to the experiment than the other model, we calculate D = D_con − D_int, shown in the table. The columns R_95% provide Bonferroni-corrected 95% confidence radii on D. If the interval [D − R_95%, D + R_95%] does not include zero, we conclude that one model produces data more faithful to the experiment than the other model. Cases where the confidence interval excludes zero are colored green or red depending on whether the interactive model or control model, respectively, is more faithful to experiment.