. 2010 Sep 8;5(9):e12633. doi: 10.1371/journal.pone.0012633

Table 4. Sensitivity analysis of the associations between ASD status and the 7 Factors or the 4 individual measures.

Trait	All Data				At least half observed				Complete data
	OR	95% CI		p	OR	95% CI		p	OR	95% CI		p
1: Verbal ability	1.38	1.26	1.51	<0.001	1.48	1.34	1.65	<0.001	0.92	0.43	2.00	0.837
2: Language acquisition	1.65	1.26	2.17	<0.001	1.78	1.28	2.49	0.001	0.59	0.19	1.78	0.345
3: Social understanding	2.27	1.88	2.73	<0.001	2.41	1.93	3.00	<0.001	2.85	1.39	5.86	0.004
4: Semantic-pragmatic skills	1.88	1.55	2.28	<0.001	1.95	1.57	2.43	<0.001	2.11	0.77	5.74	0.144
5: Repetitive-stereotyped	1.32	1.20	1.45	<0.001	1.32	1.18	1.47	<0.001	1.42	0.87	2.31	0.159
6: Articulation	1.41	1.16	1.70	<0.001	1.58	1.27	1.95	<0.001	2.12	0.94	4.78	0.072
7: Social inhibition	1.67	1.36	2.06	<0.001	1.61	1.28	2.02	<0.001	1.32	0.68	2.55	0.410
N	13138				9375				2481
ASD cases	80				61				4
Increase in R²	0.450				0.501				0.399
CCC – coherence 9y	1.86	1.63	2.14	<0.001	1.94	1.67	2.25	<0.001	2.18	1.81	2.63	<0.001
SCDC 91 m	1.58	1.34	1.87	<0.001	1.58	1.32	1.89	<0.001	1.62	1.31	2.01	<0.001
Repetitive behaviour 69 m	1.16	1.04	1.30	0.011	1.15	1.02	1.30	0.026	1.09	0.91	1.29	0.345
EAS – sociability 38 m	1.77	1.42	2.21	<0.001	1.79	1.41	2.29	<0.001	1.61	1.19	2.18	0.002
N	11411				9422				6055
ASD cases	73				63				44
Increase in R²	0.482				0.527				0.544

See Methods S3 for definitions of the abbreviations associated with the individual measures. Logistic regression was used to test the association between the traits and ASD status adjusting for gender. Traits were included as linear covariates. ORs are for one SD decrease in trait score. All data consists of at least one individual measure with observed data (from the 93 measures used to derive the factors or from the 4 measures included in the individual measure model) with missing data being imputed. Complete data consists of observed data only. R² are reported as the increase in the explanation of the log-likelihood compared to a model involving gender only. Gender explained about 5% of the log likelihood in all models. The estimated relative contribution of each trait using All data were: 19%, 5%, 31%, 17%, 13%, 5% and 10% (Factors); and, 57%, 19%, 4% and 19% (individual measures). The factor model performed similarly to the individual measure model when restricted to the same sample (increase in R² = 0.493, N = 11411).