91Ó°ÊÓ

When evaluating the efficiency of a regression model, the Coefficient of Determination, commonly denoted as \( R^2 \), plays a crucial role. This coefficient tells us the proportion of the variance in the dependent variable, in this case, the lift/drag ratio \( y \), that can be explained by the independent variables in the model. In simpler terms, \( R^2 \) measures how well our chosen variables can predict the behavior of \( y \).

To calculate \( R^2 \), we use the formula: \[ R^2 = 1 - \frac{\text{SSE}}{\text{SST}} \] where SST is the total sum of squares, representing the total variation in \( y \), and SSE is the sum of squared errors, which accounts for the unexplained variance by the model.

For the first-order model (without interaction), with SSE of 5.18 and SST of 8.55, the \( R^2 \) comes out to be 0.394. This means that the model explains approximately 39.4% of the variability in the lift/drag ratio, leaving the rest unexplained. When we add the interaction term \( x_3 = x_1x_2 \), the SSE drops to 3.07, increasing \( R^2 \) to 0.641. Hence, the interaction model explains 64.1% of the variance, showing a significant improvement in prediction power.

The larger the value of \( R^2 \), the better the model fits the data. In this case, adding the interaction term greatly enhances the model's explanatory power.

An interaction model in regression analysis incorporates both the main effects of the predictors and the interaction between them. The term interaction refers to the scenario where the effect of one independent variable on the dependent variable changes depending on the level of another independent variable.

In our aeronautical engineering experiment, the interaction model includes an additional term \( x_3 = x_1 x_2 \). This interaction term is crucial because it captures the combined effects of the position of the forward lifting surface and the tail placement relative to the main wing. By considering the interaction between these two variables, we can uncover relationships that aren't apparent when observing them independently.

Using an interaction model is particularly useful when you suspect that the relationship between the predictors and the outcome variable is not merely additive. For instance, the effect of the tail placement \( x_2 \) on lift/drag ratio \( y \) might be stronger when the forward lifting surface \( x_1 \) is at a specific position. Incorporating \( x_3 \) helps in better capturing such nuances.

In this example, introducing the interaction term reduced the sum of squared errors from 5.18 to 3.07, evidenced by a higher \( R^2 \) in the interaction model. This clearly illustrates the value and insight gained by modeling these variables together rather than in isolation.

The model utility test is used to determine whether a regression model provides significant predictions for the dependent variable from the independent variables. Essentially, this test helps us confirm if the model is useful.

To perform this test, we use an F-statistic calculated from the ratio of explained variance by the model to the unexplained variance. This formula is: \[ F = \frac{\frac{(\text{SST} - \text{SSE})}{p}}{\frac{\text{SSE}}{n-p-1}} \] where \( p \) is the number of predictors, and \( n \) is the total number of observations.

For the first-order model, the F-statistic calculated was 2.958, which is less than the critical F-value of 5.14 for a significance level \( \alpha = 0.05 \). This result suggests that the predictors in the first-order model do not provide a significant explanation of the variance in lift/drag ratio. In other words, the model might not be very useful in its current form.

Conversely, the interaction model significantly improves the prediction capability, with an F-statistic of 5.237, surpassing the critical value of 4.76. Therefore, the interaction model is considered significant and useful for explaining the variability in the data.

In summary, a model utility test is crucial in regression analysis to gauge whether the models are effective, highlighting the importance and value added by including an interaction term in this particular analysis.