Table 3.

Structural equation model for fat mass gain during feeding trial and serum HDL cholesterol concentration after the feeding trial.

Conditioned and unconditioned QTL scans
Phenotype	Model	Condition	QTL	LOD (−log10(P value))ξ
Fat mass gain (during feeding trial) (g)	Combined (both sexes, both diets) (y ~ sex * diet + marker)	Unconditioned	Fmgq1	4.02*
	Combined (both sexes, both diets) (y ~ sex * diet + m + HDL)	Serum HDL cholesterol concentration (after feeding trial) (ng/mL)	Fmgq1	4.58*
Serum HDL concentration (after feeding trial) (ng/mL)	Combined (both sexes, both diets) (y ~ sex * diet + marker)	Unconditioned	Hdlq1	11.94
	Combined (both sexes, both diets) (y ~ sex * diet + marker + fat mass gain)	Fat mass gain (during feeding trial) (g)	Hdlq1	10.86
Causal network analysis
Relationship	Model	P value	AIC	BIC
Fat mass gain ← serum HDL cholesterol	(Fat mass gain ~ HDL) ~ sex * diet + top marker at Fmgq1	0.0046a	2181	2209
Fat mass gain → serum HDL cholesterol	(HDL ~ fat mass gain) ~ sex * diet + top marker at Fmgq1	0.0252a	4204	4232
Structural model
	Variable	Predictor	Path coefficient (% variation explained)	t-statistic of path coefficient
	Fat mass gain (during feeding trial) (g)	Sex	0.43 (17.19%)	9.89
		Sex * diet	−0.09 (1.02%)	−2.03
		Fmgq1	0.19 (3.99%)	4.40
		Fmgq2	−0.21 (5.19%)	−4.87
		Fmgq3	−0.11 (2.56%)	−2.55
	Serum HDL cholesterol concentration (after feeding trial) (ng/mL)	Sex	0.44 (17.37%)	10.09
		Sex * diet	−0.01 (0.04%)	−2.03
		Hdlq1	0.31 (9.59%)	7.20

^ξLOD (−log10(P values)) is provided for the top marker in each confidence interval.

^aCausal inference is significant at Holm–Bonferroni corrected P < 0.05.