REGRESSION MODELS FOR NON-MOTION SIMULATORS
Regression Model Analysis
The initial scatterplot for the number of non-motion simulators and the number of corresponding helicopters showed that cubic and quadratic equations would describe the regression model with greater precision than the rest.
The high number of the U.S. Army Black Hawks and the relatively low number of corresponding non-motion devices, though, proved to be misleading. The models did not have a logical practical explanation. For example, the quadratic and cubic models predicted that 1700 helicopters do not need any simulators at all. Thus, this extreme value had been excluded and the scatterplot was run again.
This time the scatterplot showed that linear, cubic and quadratic equations would describe the regression model with greater precision than the rest. Also, these models seemed very close to each other.
The SPSS “curve estimation” analysis for the linear equation gave the results shown on tables 11, 12, and 13.
The SPSS “curve estimation” analysis for the quadratic equation gave the results shown on tables 14, 15, and 16.
The SPSS “curve estimation” analysis for the quadratic equation gave the results shown on tables 17, 18, and 19.
All three models indicated that there is a strong correlation between the number of helicopters and the corresponding non-simulators (R2=.851, and R2=.852). Figure E2 presents the curve fit of the three equations. At the scatterplot it is evident that all three equations would give similar predictions for the numbers of non-motion simulators. The linear model, though, was chosen because it was the simplest one. The equation of the model is Y=0+(.034*X).