Table 1 The parameters for the assessment of these models.

NO. Model		Type	A	N	R2X(cum)	R2Y(cum)	Q2(cum)	R2	Q2
All	M1	PCA-X	4	23	0.689		0.399
em	M2	PCA-X	2	12	0.605		0.344
em	M3	PLS-DA	4	12	0.839	0.994	0.966
em	M4	OPLS-DA	1 + 3 + 0	12	0.839	0.994	0.951	0.904	−0.617
et	M5	PCA-X	2	12	0.649		0.455
et	M6	PLS-DA	3	12	0.79	0.998	0.988
et	M7	OPLS-DA	1 + 3 + 0	12	0.822	0.999	0.979	0.911	−0.27
mt	M8	PCA-X	2	12	0.522		−0.00798
mt	M9	PLS-DA	3	12	0.667	0.992	0.921
mt	M10	OPLS-DA	1 + 3 + 0	12	0.758	0.998	0.889	0.965	−0.169

Note: A stands for the PC numbers while each model is constructed, N for the numbers of samples analyzed, em for the comparative group E/M, et for the comparative group E/T, mt for the comparative group M/T, M1-M10 for Model1–10, R2X(cum) for the interpretation rate of each model in the X axis direction in multivariate statistical analysis modeling, R2Y(cum) for the interpretation rate of each model in Y axis direction in multivariate statistical analysis modeling, Q2(cum) for the prediction rate of each model, R2 for the intercept value of the Y axis and the regression line, which is obtained when Linear regression analysis between the Y matrix of the original classification, the Y matrices of N times′ different permutations and R2Y was conducted during model validation, and Q2 for the intercept value of the Y axis and the regression line, which is obtained when Linearregression analysis between the Y matrix of the original classification, the Y matrices of N times′ different permutations and Q2Y was conducted during model validation; for Q2 in external validation, general requirement is that Q2 < 0, overfitting is avoided. For R2 in internal validation, general requirement is that R2 > 0.5, the closer to 1R2, the better the model. For R2X, general requirement is that R2X > 0.4, the model is good.

Quick links

Search