. 2020 Nov 30;3(11):e2025454. doi: 10.1001/jamanetworkopen.2020.25454

Table 2. Relationship Between Changes in Thermic Effect of Food and the First Predictive Component as Evaluated by the OPLS Model.

Variable	OPLS predictive component				Multiple regression
Variable	Component loading^a	t Statistic	R^b	P value for R	Regression coefficient	t Statistic
Matrix X
Baseline BMI	0.191	2.46	0.209	<.05	−0.015	−0.33
Baseline fat mass	0.256	2.89	0.283	<.05	−0.014	−0.28
Baseline TEF	−0.850	−11.96	−0.938	.005	−0.505	−5.69^c
Change in PREDIM	0.324	2.41	0.359	<.05	0.105	1.37
Change in fat mass	−0.271	−2.59	−0.301	<.05	−0.122	−1.55
Matrix Y
Change in TEF	1.000	5.27	0.540	.003	NA	NA

Abbreviations: BMI, body mass index; NA, not applicable; OPLS, orthogonal projections to latent structure; PREDIM, predicted insulin sensitivity index; TEF, thermic effect of food.

^{^a}

The explained variability was 29.2% (24.3% after cross-validation).

^{^b}

Component loadings expressed as a correlation coefficients with predictive component.