Regression modeling approach
The data set was parsed into training and testing partitions using
methods of randomization. The validity of the model was based on
goodness-of-fit of R Sq. > 90% and ANOVA, [p value
<0.05] and a consequent F Ratio (Table 2a and 2b).These statistical results confirmed acceptable degree of predictability
of the model.
Following this multiple-regression analysis, we conducted 2,000
bootstrap samplings using the predicted coefficients and random variates
from chosen intervals of parameters. The assumption for this analysis
was that each of the parameters were independent variables. The
coefficients of each parameter were determined by using multiple
regression analyses, which is the multiplier to the parameter value in a
linear regression equation. The inclusion of all the variables in
analysis ensures their contribution to the model [41]. However,
analysts applying this model in the future may, at their judgment,
evaluate statistical significance of regression coefficients. Parameters
that are not significant maybe excluded using step wise regression. In
our analysis, results based on training dataset predictors matched with
those from the test dataset confirming an acceptable degree of
predictability of the model. We invite the readers of this article to
contact us to analyze the predictive potential of the model using their
data.