Theoretical approaches to drug design and biological activity: II. Multivariate structure‐activity relationships in combination with the multivariate bioassay (MASCA Model)

The role of diagnostic statistics in medicinal chemistry

“…confidence intervals of regression coefficients, and misleading predictions. [12][13][14][15][16] If alternative regression methods are applied, the estimates are always biased.…”

“…[12][13][14][15][16][23][24][25] (1) A multicollinearity which contributes significantly to the predictive model power of a QSAR model (type 1 multicollinearity); (2) a multicollinearity without predictive capability (type 2 multicollinearity); and (3) a mixed-type multicollinearity, that is, a "borderline" between type 1 and type 2 (type 3 multicollinearity).…”

“…Nonleast-squares (NLS) regression 21,22 is the reduced-rank version of a modified PCRA analysis, [12][13][14][15][16] combined with a retransformation of the components of the descriptors in original terms.…”

Non‐Least‐Squares Regression Analysis Applied to Organic and Medicinal Chemistry

“…Thus the tlk values are the t statistics of p x c regression coefficients blk of B. In other words, the matrix TS contains the same values as the known univariate t tests of p multiple regression equations, while the only difference is due to co. 2 It can easily be derived that the confidence intervals are defined by…”

“…Print results of (2) Transform the PCRA into the NLS regression equation -Save results (e.g., the residuals) for subsequent subroutines -End to give the intuition of researchers more play; its heuristic value cannot be underestimated. In some cases, the traditional and computer-aided drug design are much more an art than an "automatic working procedure" of correlations between activities and structure parameters.…”