Inferring Depression and Its Semantic Underpinnings from Simple Lexical Choices

Table 1

Study 1: classification performance metrics.

Model	PCs	Accuracy		95% CI				ROC AUC		Sensitivity		Specificity
		Train	Test	Train	Test	Train	Test	Train	Test	Train	Test	Train	Test
mDCT	18	.61	.65	[.56, .67]	[.56, .72]	<.001	<.001	.67	.67	.61	.67	.62	.63
mDCT+Demo	10	.62	.66	[.57, .67]	[.57, .74]	<.001	<.001	.66	.67	.61	.69	.63	.63
mGenderAge	—	.58	.56	[.52, .63]	[.48, .65]	<.05	.07	.61	.59	.53	.54	.61	.59
mRandomBaseline	35	.65	.54	[.60, .70]	[.46, .62]	<.001	.18	.70	.49	.63	.56	.68	.53

The best performing model is highlighted in bold.