Additivity models, selection

In addition, model selection should be limited in order to protect against Type I error of BE testing. Therefore, we argue that a proper assessment framework should incorporate the following features ... 

On the basis of their evaluation and our internal predictive VolSurf model [160] for this series (r 0.81, q 0.60, 4 PLS components), it can be concluded that factors like size and shape, which had previously been reported to affect paracellular permeability, are indeed important in the VolSurf PLS model to explain the local structure-permeability relationship of one particular scaffold. Hence, local statistical models provide a qualitative ranking of candidates, and thus are valuable for optimization of pharmaceutically relevant compounds, especially if combined with additional models to understand affinity, selectivity or any particular pharmacokinetic behavior. 

Validation of QSAR models is one of the most critical problems of QSAR. Recently, we have extended our requirements for the validation of multiple QSAR models selected by acceptable statistics criteria of prediction for the test set (19). Additional studies on this critical component of QSAR modeling should establish reliable and commonly accepted good practices for model development, which should make models increasingly useful for virtual screening. 

Model selection will have a profound effect on low-dose risks, and indications of this variability should be provided. Biological variability must also be described. It is essential to differentiate all of these assumptions and judgments from scientific fact. Indeed, clear distinctions should be drawn between chemicals for which qualitative evidence is overwhelming and those for which the evidence is marginal for additional insight on the quantitative conclusions. 

The first step is to determine whether a classical additive model, an outranking approach or DEA should be used. The criteria to be used for the site selection/site ranking task do not lend themselves to a clear classification into input and output parameters. Furthermore, the relatively complex mathematical procedure underlying DEA was perceived to be inappropriate for support of top management decisions that are to a large extent driven by qualitative factors. Hence, DEA was ruled out first. 

It is possible to improve the method by scaling the data, but it is important to be very careful to think about the consequences of the various methods employed. It is sometimes possible to scale first the two way data and then unfold. However, a final centring should normally be performed on the unfolded matrix. In addition, variable selection can have a significant influence on the effectiveness of unfolded PLS models, since not all the 2040 variables are going to be particularly relevant or informative. 

Hence, local statistical models provide a qualitative ranking of candidates and, thus, are valuable for optimization of pharmaceutically relevant compounds, especially if combined with additional models to understand affinity, selectivity, or any particular pharmacokinetic behavior. 

One method used adsorption to laboratory surrogate materials in an adsorption additivity model to estimate the relative contributions of Mn and Fe oxides and organic materials to Pb adsorption to natural biohlms. The second method made use of a novel selective extraction approach in which trace metal adsorption to surface coatings was measured before and after the selective removal of constituents. 

Several related problems have been previously considered in the literature. In addition to the afore mentioned statistical approaches for structural change detection in data sets and their application for linear system identification [7], the joint problem of model structure determination and parameter estimation was addressed by [8-10]. A related approach was used by [11-13] in the context of data reconciliation. Additional aspects of model selection in chemical engineering are covered in [14]. Although the present problem shares common features with the all of the previous applications, it also presents unique characteristics that require a specific formulation. 

Additional parameters arise according to the model selected and hence the physical effects that are taken into account. Parameters appropriate for the general rate model are given by Berninger et al. (1991) and Ma et al. (1996). 

Exploratory modeling using modem statistical modeling techniques such as generalized additive modeling (GAM) (15), cluster analysis, and tree-based modehng (TBM) to reveal structure in the data and initially select explanatory covariates. 

---