Predictors of first-year GPA of medical students: a longitudinal study of 1285 matriculates in China

BMC Medical Education

Table 5 Variance component of random effects by ANOVA and proportional reductions in residual variance for conditional models

	Fully unconditional model	Conditional model (hierarchical linear model)*
	Fully unconditional model	Model 1		Model 2		Model 3
Random effects	Variance component (%)^†	Variance component	Proportion reduction^§	Variance component	Proportion reduction^§	Variance component	Proportion reduction^§
Between-major residual	4.05 (14.8)	1.63	59.9%	1.63	59.7%	1.76	56.6%
Within-major residual	23.30 (85.2)	23.31	0.0%	21.93	5.9%	20.63	11.5%
Total residual	27.35 (100)	24.93	8.9%	23.57	13.8%	22.39	18.2%

*Conditional model: Model 1 included major-level variables alone; Model 2 included major-level variables and student-level socio-demographic variables; Model 3 included all variables.
^†"%" denotes the proportion of total variance due to between- and within-major variance, correspondingly.
^§"Proportion reduction" denotes the proportion reduction in residual variance (δ²), calculated by [δ²(unconditional)-δ²(conditional)]/δ²(unconditional).

ISSN: 1472-6920