(21) The corresponding

R-squared values, or explanatory power, of the models in Table 3 are lower than in Table 2, and range from 0.0818 to 0.2410.

This is why the single or multivariate models with or without change-points (Models 1, 2, and 3) produce very poor results with the very low

R-squared values shown in Table 2, specifically for Buildings 1, 2, and 3 that have more different data clusters.

R-Squared Values for the Proposed Prediction Models for Building #1 Models Greensboro Seattle Phoenix Miami Linear Regression 0.7495 0.6357 0.8929 0.9343 Neural Network 0.9960 0.9972 0.9874 0.9647 Table 3.

[5] 16.4 -- -- -- Table 3: Adjusted

R-squared values (%) of the response functions.

Input: Stipulated number of iterations I, current number of iterations i, previous retrieval equation [y.sub.1](x), current retrieval equation [y.sub.2](x), and corresponding

R-squared values [R.sub.1] and [R.sub.2] Output: Target retrieval equation y(x) (1) Designate retrieval weight, with 1 g as sampling point, retrieval from 5 g to 15 g; (2) Calculate retrieval equation [y.sub.1](x) using LSR (Eq.

Bivariate

r-squared values were calculated using Statistix 9 analytical software (Statistical Software, Tallahassee, FL) for indicating the predictive power of the independent variables relative to the level of injury from an H-ARS RC3 condition.

Table 5 summarizes the full model, which resulted in an

R-Squared value of .087.

R-squared values for the load-displacement curves from experiment and the simulations adopting different stress-strain curves.

However, it is worth noting that the

R-Squared value obtained when using actual hurricane variable (data) is significantly higher when compared with the

R-Squared value obtained using the predicted hurricane variable, which appears in Table 5 and Table 6.

The

R-squared value is not very high for the model (60%) due to characteristics of the sample however; regression model is highly significant (Probability .

A trendline is the most reliable when its

R-squared value is at or near 1.

In each case, third order polynomial trend lines were used to fit the data, and the

R-squared value for the fit is indicated on the graphs.