Screening variables

There is a spectrum of methods proposed for screening variables in a computer experiment. They differ mainly in the assumptions they make about the form of an input-output relationship with stronger assumptions, fewer runs are typically required. 

The process control display is divided into three sections. Six process control variable displays are shown at the top of the screen. Variable displays reflect the most recent measurements of the particulate. A time... 

The basic approach was to set up a large number and wide range of simulated correlation studies using random numbers for both observations and screened variables and assess the incidence and degree of any resulting correlations. 

From data generated in the present study it is now possible to answer the question raised earlier concerning the p-value of equation 1. With ten observations and eight screened variables the expected frequency of chance correlations comes out to be about 12. Thus, the validity of the correlation equation is far less certain than the reported p-value indicates. 

In Figure 2 the relationship between the number of observations and the probability of a chance correlation with ten screened variables for various r2 values is shown. From the graph corresponding to r2 0.8 it may be seen that the probability of encountering a chance correlation at this level is about 22 for ten observations, reaching zero for 23 observations. 

Figure 2. Relationship between number of observations and probability of a chance correlation (Pc) for 10 screened variables with r2 0.7 (O), 0.8 (A), and...

Extrapolation of the previous graph allows one (Figure 6) to make similar estimations for larger numbers of screened variables. Thus, some 56 observations would be required to screen kO variables while keeping the probability of encountering a chance correlation with r2 0.8 at the 1 level or less. 

In the light of the chance correlation phenomenon described, some comments will be made on a number of correlation studies reported in the literature which involved large numbers of screened variables. 

The present study provides guidelines for the approximate incidence of chance correlations at specified r2 values for various combinations of observations and screened variables. Specific correlations obtained under conditions where the guidelines indicate an appreciable risk of chance correlations should be individually checked as described. 

FI0.6. 11. Plot of mean versus the number of screened variables for regression equations generated for sets of random numbers (from Topliss and Edwards 1979, copyright (1979)... 

Adsorption is a more fundamental property of the polymer-rock surface-solvent system and cannot be avoided in the same way except by trying another polymer. Thus, adsorption is the most important mechanism that should be studied in evaluating a given polymer flood application. Retention via mechanical entrapment should be treated as a screening variable and should be avoided. Hydrodynamic retention is generally small and can be neglected in most practical applications. Because of these views, most of the rest of this chapter will concentrate mainly on adsorption. This approach... 

---