TABLE 5.

Best-fitting regression models for tactile detection thresholds.

Whole sample: best-fitting model (Baseline + c)				Adults: best-fitting model (Baseline + c)
Predictor	aOR (95% CI)	Wald χ2	P	Predictor	aOR (95% CI)	Wald χ2	P
Diagnosis (ASD)	1.48 (0.82, 2.66)	1.67	0.196	Diagnosis (ASD)	0.96 (0.38, 2.44)	0.01	0.933
Sex (male)	2.06 (1.06, 4.0)	4.6	0.032	Sex (Male)	2.15 (0.8, 5.8)	2.31	0.129
Age (years)	1.06 (1.03, 1.1)	14.78	<0.001*	Age (years)	1.09 (1.01, 1.18)	5.27	0.022*
Counterbalance	0.71 (0.4, 1.28)	1.31	0.253	Counterbalance	0.43 (0.15, 1.18)	2.69	0.101
C	76.38 (20.78, 280.7)	42.63	<0.001*	c	39.28 (4.89, 315.6)	11.92	<0.001
Model fit	χ2(5) = 77.12	p < 0.001*	R2 = 0.408	Model fit	χ2(5) = 25.10	p < 0.001*	R2 = 0.356
Children: best-fitting model (Baseline + c)
Predictor	aOR (95% CI)	Wald χ2	P
Diagnosis (ASD)	2.12 (0.93, 4.85)	3.18	0.075
Sex (Male)	2.56 (0.93, 7.03)	3.34	0.067
Age (years)	0.94 (0.83, 1.06)	1.1	0.295
Counterbalance	0.86 (0.39, 1.88)	0.15	0.701
c	142.8 (24.11, 845.6)	29.88	<0.001
Model fit	χ2(4) = 49.52	p < 0.001*	R2 = 0.423

Significant predictors in each model are bolded; aOR, Adjusted Odds Ratio. Best fitting model was determined using a best subset regression procedure and Bayesian Information Criterion (BIC). c, response criterion. ^∗p < 0.05.