Latent class analysis of actigraphy within the depression early warning (DEW) longitudinal clinical youth cohort

Table 4 Detailed summary of linear mixed-effects model output

Outcome	Predictor¹	\(\:\beta\:\) estimate	Standard error	Degrees of freedom	t-value	95% CI	p-value	Effect size
Acceleration⁴	Intercept	64.202	8.821	3119	7.278	(47.006, 81.398)	< 0.001
	Male vs. female	2.894	2.661	56	1.088	(−2.406, 8.194)	0.281	0.176
	Age	−1.322	0.469	56	−2.816	(−2.257, −0.387)	0.007	−0.081
	PHQ-9	−4.827	2.479	56	−1.947	(−9.765, 0.111)	0.057	−0.294
	RRS	−0.109	0.316	56	−0.343	(−0.739, 0.522)	0.733	−0.007
	Arm 2 vs. arm 1²	−2.516	0.688	3119	−3.66	(−3.857, −1.176)	< 0.001
	Arm 3 vs. arm 1	−0.375	0.775	3119	−0.484	(−1.886, 1.136)	0.629
	Arm 4 vs. arm 1	−2.555	0.931	3119	−2.745	(−4.369, −0.74)	0.006
	Arm 5 vs. arm 1	−7.112	1.608	3119	−4.423	(−10.247, −3.977)	< 0.001
Sleep duration⁵	Intercept	477.712	64.552	3122	7.4	(351.815, 603.608)	< 0.001
	Male vs. female	−2.244	3.435	56	−0.653	(−9.088, 4.6)	0.516	−0.02
	Age	20.474	19.486	56	1.051	(−18.355, 59.302)	0.298	0.182
	PHQ-9	40.953	18.125	56	2.259	(4.837, 77.07)	0.028	0.364
	RRS	−2.586	2.314	56	−1.117	(−7.198, 2.025)	0.269	−0.023
	Arm³	1.756	1.709	3122	1.028	(−1.577, 5.089)	0.304

¹The night variable (30 levels) was included in the model as a categorical predictor but their estimates are not included here
²Arm was treated as a categorical predictor since it improves the fit significantly (p-value < 0.001) in contrast to being treated as a semi-continuous predictor
³Arm was treated as a semi-continuous predictor since being treated as a categorical predictor did not improve the fit significantly (p-value = 0.292)
⁴Model diagnostic analysis did not show deviation from the normality assumption (p-value = 0.353, Kolmogorov-Smirnov Test). About 5.4% of the observations were categorized as high influential cases. However, a sensitivity analysis with these cases deleted did not show substantial differences
⁵Model diagnostic analysis showed deviation from the normality assumption (p-value = 0.002, Kolmogorov-Smirnov Test). Given the large sample size, this has little impact on the validity of the tests for significance. About 5.3% of the observations were categorized as high influential cases. A sensitivity analysis with these cases deleted changed the significance of PHQ-9 with a p-value of 0.061

ISSN: 1753-2000