An overlooked property of plot methods

被引：13

作者：

Besalu, Emili

de Julian-Ortiz, J. Vicente

Iglesias, Monica

Pogliani, Lionello

机构：

[1] Univ Girona, Fac Ciencies, Inst Computat Chem, Girona 17071, Spain

[2] Univ Valencia, Fac Farm, Unidad Invest Deseno Farm & Conectividad Mol, Dept Quim Fis, Valencia 46100, Spain

[3] Red Invest Ctr Enfermedades Trop, Dept Biol Celular & Parasitol, Valencia 46100, Spain

[4] Univ Girona, Dept Chem, Girona 17071, Spain

[5] Univ Calabria, Dipartimento Chim, I-87030 Arcavacata Di Rende, CS, Italy

来源：

JOURNAL OF MATHEMATICAL CHEMISTRY | 2006年 / 39卷 / 3-4期

关键词：

multilinear regression properties; plot properties; error plots; calibration; MLR theorem;

D O I：

10.1007/s10910-005-9035-z

中图分类号：

O6 [化学];

学科分类号：

0703 ;

摘要：

An interesting and often overlooked property of the MLR least-squares method is discussed. Here it is shown how the observed vs. calculated and calculated vs. observed plots, which are usually presented as equivalent, in fact they are not. Due to the inherent properties of the MLR procedure, it is shown that the slope of the calculated vs. observed plot is r(2), as the slope of the experimental vs. calculated plot is just 1. This behaviour also has direct consequences to the corresponding residual plots.

引用

页码：475 / 484

页数：10

共 28 条

[1] Global and local computational models for aqueous solubility prediction of drug-like molecules [J].