Variable screening model

Figure 23. Radial coupling matrix element for the 3o- and 4a MO s in the system Ne + Kr. The curve labeled SHM refers to calculations using the model matrix elements (18) and the circles labeled VSM are calculated by Fritsch and Wille on the basis of the variable screening model/ ...

Figure 24. Excitation probability of the Kx L shell in Ne + Kr collisions versus the inverse projectile velocity. The curve labeled NM(2) follows from two-state calculations using Nikitin s model. The curves labeled SHM(2) and SHM(3) represent model calculations using matrix elements (18) for two and three states, respectively. The curve labeled VSM(3) refers to three-state calculations by Fritsch and Wille using the variable screening model. ...

An experiment involving a complex computer model or code may have tens or even hundreds of input variables and, hence, the identification of the more important variables (screening) is often crucial. Methods are described for decomposing a complex input-output relationship into effects. Effects are more easily understood because each is due to only one or a small number of input variables. They can be assessed for importance either visually or via a functional analysis of variance. Effects are estimated from flexible approximations to the input-output relationships of the computer model. This allows complex nonlinear and interaction relationships to be identified. The methodology is demonstrated on a computer model of the relationship between environmental policy and the world economy. 

In applications such as the Arctic sea ice model (Chapman et al., 1994) mentioned above, a strategic objective of a preliminary computer experiment is screening finding the important input variables. Screening is not a trivial task because the computer model is typically complex, and the relationships between input variables and output variables are not obvious. A common approach is to approximate the relationship by a statistical surrogate model, which is easier to explore. This is particularly useful when there are many input variables. 

The Plackett-Burman designs are convenient for fitting linear screening models when the number of variables is large and when it is desirable to keep the number of necessary runs to a minimum. One disadvantage is that the confounding patterns... 

However, the limited test of the screening model mobility prediction is more encouraging. Given that a certain amount of variability will occur under field conditions, the laboratory R values did present a picture of the relative order of leaching of the four chemicals in the two plot experiments. If this result is obtained when the field experiment is repeated on large numbers of chemicals in different soils, it may eventuality allow future comprehensive field research studies to focus on representatives of different mobility categories rather than requiring prohibitive numbers of experiments to be conducted on each chemcial. 

Consider the screening problem of table 2.20. Here we want to know the effect of various diluents, disintegrants, lubricants etc. on the stability of the drug substance. We have not imposed any restraint whatsoever on the numbers of levels for each variable. We could of course test all possible combinations with the full factorial design, which consists of 384 experiments (4 x 2 x 3 x 4 x 2 x 2) Reference to the general additive screening model for different numbers of levels will show that the model contains 12 independent terms. Therefore the minimum number of experiments needed is also 12. 

This is a screening model, an additive (first-order) model for 6 factors, where the S levels of the / qualitative variable are replaced by s, - 1 independent (presence-absence) variables ... 

In the near future, more sophisticated models can be built using probabilistic networks. A probabilistic network is a factorization of the joint probabiHty function over all the considered variables (markers, interventions, and outcomes) based on knowledge about the dependencies and independencies between the variables. Such knowledge is naturally provided by the hits coming out of the association screen, where each association can be interpreted as a dependency, and the absence of an association as an independency between variables. The model can then be parameterized by fitting to the data, similarly to the linear and logistic regression models, which are in fact special cases of probabilistic network models. 

The PLOT statement usually appears in the DYNAMIC part of the model and controls the output to the screen during a run after a START command. Only two variables may be specified. The format of this command is PLOT X variable, y variable, xmin, xmax, ymin, ymax... 

Screening designs are mainly used in the intial exploratory phase to identify the most important variables governing the system performance. Once all the important parameters have been identified and it is anticipated that the linear model in Eqn (2) is inadequate to model the experimental data, then second-order polynomials are commonly used to extend the linear model. These models take the form of Eqn (3), where (3j are the coefficients for the squared terms in the model and 3-way and higher-order interactions are excluded. 

A complete list of the reaction conditions tested for this response surface design can be found in [76], The center point reaction condition was repeated six times. This was done to measure the variability of the reaction system. Also, the space velocity is kept constant, as it was the least important factor predicted by screening design, for all the reaction conditions. The purpose of this nested response surface design was to develop an empirical model in the form of Eqn (5) to relate the five reaction condition variables and the three catalyst composition variables to the observed catalytic performance. 

Two main factors have guided the need for optimization of the early screening techniques on one hand the use of simple, quick and high-capacity cell monolayer methods, e.g., Caco-2 cell and MDCK and on the other hand the increased synthesis of more lipophilic, insoluble compounds from combinatorial libraries. This has created a vast number of different variants of cell-based assays and has resulted in variability among the data obtained. A need for optimization of as many as possible of the different parameters in order to increase the predictivity and throughput of the model has been suggested in the literature [98-100]. 

---