Page 286 - Solid Waste Analysis and Minimization a Systems Approach

264 SOLID WASTE ESTIMATION AND PREDICTION

Before conducting the stepwise regression analysis, waste groups with five or less
samples were removed from the study. This was done to properly establish relation-
ships among the variables, because sample sizes of less than five could indicate
strong relationships when they do not actually exist. Owing to less than five samples,
two waste groups were removed from the study: the forestry waste group and the
motor freight and transportation waste group. This reduced the number of waste
groups from 22 to 20.
The next section discusses the process to determine the significant independent vari-
ables that aid in the prediction of annual solid waste for each waste group.

16.3 Stepwise Regression

Methodology

Multivariable stepwise regression analysis was conducted on each of the 20 waste
group data sets to determine the significant independent variables that influence the
annual solid generation. The multivariable stepwise regression analysis of the inte-
grated environmental model functioned to predict and evaluate annual solid waste
quantity means and variances of the individual companies in each of the 20 waste
groups. Linear, nonlinear, indicator (binary), and interaction variables were consid-
ered and statistically tested as potential independent variables. These variables were
gathered during the data collection phase. A five-step procedure was utilized to con-
duct the multivariable stepwise regression analysis and each step is discussed in the
following paragraphs. The five-step regression procedure is listed below:

1 Consolidate data for each of the 20 waste groups
2 Determine the multiple regression mathematics and procedure
3 Conduct regression procedure and select independent variables
4 Evaluate results and determine final models
5 Validate regression assumptions

16.3.1 STEP 1: CONSOLIDATE DATA FOR EACH
WASTE GROUP

This step prepared the waste records in each of the 20 waste groups for the regression
analysis by separating waste records for each waste group and arranging the values in a
matrix. The rows of each matrix represented each waste record and the columns repre-
sented potential independent variables to be investigated. The value to be predicted was:

y = total annual waste generation (tons/year) for each waste group

The potential independent variables to be investigated were

x = number of employees at the facility
1
x = ISO 9000 certification duration (months)
2