Prediction of Bridge Component Ratings Using Ordinal Logistic Regression Model

Table 7

Model comparison statistics with 2014 data.

Model Statistics	Multiple Linear Regression			Ordinal Regression
	Deck	Superstructure	Substructure	Deck	Superstructure	Substructure
Sample Size	2281	2906	2906	2281	2906	2906
Sum of Absolute Residuals	1550	1793	2247	1350	1441	2129
Sum of Residual Squares	2350	2493	3841	1832	1823	3417
Bias	1.087	1.079	1.105	1.025	1.022	1.053
Accuracy	1.115	1.098	1.156	1.101	1.080	1.151
Out-of-range Forecasts	0%	0%	0.1%	0%	0%	0%
Percent Exact Forecasts	46%	48%	43%	50%	56%	44 %
Percent Estimation within 1 Rating Difference	86%	91%	84%	93%	95%	87%
Percent Estimation within 2 Rating Differences	96%	98%	94%	100%	99%	97%