Prediction of Bridge Component Ratings Using Ordinal Logistic Regression Model

Table 4

Model comparison statistics with 2012 data.

Model Statistics	Multiple Linear Regression			Ordinal Regression
	Deck	Superstructure	Substructure	Deck	Superstructure	Substructure
Sample Size	2354	2991	2991	2354	2991	2991
Sum of Absolute Residuals	1,656	1,906	2,380	1,433	1,494	2,200
Sum of Residual Squares	2,446	2,722	4,090	1,933	1,986	3,580
Bias	1.087	1.090	1.128	1.026	1.023	1.078
Accuracy	1.117	1.112	1.180	1.103	1.081	1.174
AIC	5,946	6,875	8,571	5,416	5,994	7,541
LL	-2,957	-3,421	-4,270	-2,845	-3,131	-3,950
Out-of-range Forecasts	0.38%	0.87%	1.07%	0%	0%	0%
Percent Exact Estimations	44%	47%	41%	48 %	56 %	44%
Percent Estimation within 1 Rating Difference	87%	90%	83%	92%	95%	89%
Percent Estimation within 2 Rating Differences	96%	96%	93%	99%	100%	98%