Table 3.

Regional models with fixed effects, random effects and standard errors for PCS in 2004

	Model 1 (baseline)		Model 2 (+ sociodemographic variables)		Model 3 (+ sociodemographic/ selected regional variables**)
	β (SE)	p-value	β (SE)	p-value	β (SE)	p-value
Fixed effects
Constant	49.37* (0.11)	< 0.001	51.62* (0.17)	< 0.001	51.94* (0.18)	< 0.001
Level 1 (individuals)
Age	-0.31* (0.00)	< 0.001	-0.31* (0.00)	< 0.001	-0.31* (0.00)	< 0.001
Male	0.92* (0.11)	< 0.001	0.70* (0.11)	< 0.001	0.71* (0.11)	< 0.001
Migration background			-0.19 (0.18)	0.296	-0.31 (0.18)	0.084
Education
High			Ref.		Ref.
Intermediate			-1.85* (0.17)	< 0.001	-1.87* (0.17)	< 0.001
Low			-3.10* (0.18)	< 0.001	-3.21* (0.18)	< 0.001
Still in school			-3.92* (0.47)	< 0.001	-3.93* (0.47)	< 0.001
Not specified			-3.11* (0.38)	< 0.001	-3.18* (0.38)	< 0.001
Unemployed			-0.80* (0.23)	< 0.001	-0.73* (0.23)	0.001
Level 2 (households)
Single household			0.25 (0.18)	0.150	0.23 (0.18)	0.185
Net equivalence income (log)			1.40* (0.12)	< 0.001	1.31* (0.13)	< 0.001
Level 3 (regions)
East					-1.03* (0.21)	< 0.001
Random variation
Level 1 (individuals)	57.97 (0.81)		57.63 (0.80)		57.69 (0.80)
Level 2 (households)	12.20 (0.72)		9.87 (0.69)		9.84 (0.69)
Level 3 (regions)	1.73 (0.25)		1.45 (0.23)		1.24 (0.21)
-2*loglikelihood (number of cases)	147,201 (20,775)		146,480 (20,775)		146,456 (20,775)

Notes: PCS, physical component score; β, beta coefficients; SE, standard errors. * significant at the 5% level using the Wald test. ** Regional variables were selected using backward selection based on Wald test, p-values > 0.1.