Forward Selection Predictor Variables Added into the Model

Forward Selection Predictor Variables Added into the Model 

In this procedure, x, predictor variables are added into the model, rme at a time. The predictor thought to be most important by the researcher generally is added first, followed by the second, the third, and so on. If the contribution of the predictor value is unknown, one easy way to find out is to run k simple linear regressions, selecting the largest P of the k as Xi, the second largest P as X2, and so forth. 

Let us perform the procedure using the data from Example 4.2 (Table 4.10). This was an evaluation using six x,- variables in predicting the total amount of growth medium for a continuous bioreactor—biofilm—process. The researcher ranked the predictor values in the order of perceived value Xj, temperature (°C) X2 logio microbial count per cm coupon X3, medium concentration X4, calcium/phosphorous ratio X5, nitrogen level, and Xg, heavy metals. 

Because Xi is thought to be the most important predictor x, value, it is added first. We use the six-step procedure for the model-building process. 

Despite the surprising result that temperature has little direct effect on the medium requirements, the researcher moves on, using data for X2 (logio microbial count) in the model. 

---