Table 3.

Relationships between ∆BMI and relevant parameters as evaluated by the OPLS model and multiple regression analyses

		OPLS				Multiple regression
	Variable	Component loading	t-statistics	Ra		Regression coefficient	t-statistics
Relevant predictors (matrix X)	∆EDE-Q eating concern	−0.297	−3.04	−0.469	**	−0.089	−1.73
	∆EDE-Q weight concern	−0.23	−2.41	−0.364	*	0.023	0.3
	Hospitalization (days)	0.525	13.79	0.83	**	0.372	3.83	**
	Adulthood stress	−0.285	−8.05	−0.451	**	−0.131	−1.89
	∆Gamma globulin (%)	0.257	1.97	0.406	*	0.078	1.25
	∆Gamma globulin (g/l)	0.241	2.26	0.381	*	0.088	1.63
	∆IgM (g/l)	0.226	2.66	0.357	*	0.15	2.43	*
	∆fT4 (pmol/l)	−0.217	−2.53	−0.343	*	−0.167	−2.09	*
	∆AN DSM mild	0.315	2.88	0.498	*	0.289	2.78	*
	∆AN DSM extreme	−0.324	−4.74	−0.513	**	−0.225	−3.9	**
(matrix Y)	∆BMI (kg/m2)	1	20.95	0.896	**
Explained variability		80.2% (69.3% after cross-validation)

^aR Component loadings expressed as correlation coefficients with predictive component, *p < 0.05, **p < 0.01; fT4 – free thyroxine