Table 2

Multivariate predictors of annual Harris County single-family submersions Poisson-Lognormal-CAR model coefficients and 95% credible intervals (N=786 census tracts)

DepVar: log of annual number of single-family submersions per 100 per tract
Model 1Model 2
N786786
Df775779
Number of samples200 000200 000
Number of ‘burn in’ samples100 000100 000
Log likelihood−8366.1−8130.6
AIC16 754.316 275.2
BIC16 805.616 307.9
Deviance15 589.5****15 123.8****
Mean absolute deviation4.343.08
Mean-squared error3558.9612.7
Dispersion multiplier45.2****43.6****
Coefficient(95% credible interval)Coefficient(95% credible interval)
Modelled parameters
Constant−16.5050(−23.9792 to −11.5879)−13.1852(−16.1096 to −10.4492)
Socioeconomic variables
African-American population0.0004(−0.0007 to  0.0013)
Asian population−0.0003 (−0.0030 to 0.0021)
Hispanic population0.0008 (−0.0001 to 0.0017)
Median household income ($'000)−0.00003(−0.00005 to 0.0001)
Exposure variables
Number of single-family pools 0.0131(0.0060 to 0.0208)0.0120(0.0069 to 0.0175)
Number of children, ages 0–17, in single-family residences−0.0007(−0.0005 to 0.0020)0.0014(0.0004 to 0.0024)
Statistical controls
Area of tract (m2)−0.1674(−0.4486 to 0.0885)−0.1633(−0.4545 to 0.0785)
Distance to city centre (m)0.1792(0.0161 to 0.3515)0.1170(−0.0413 to 0.2669)
Unmeasured spatial effect
Spatial autocorrelation (Φ)−0.0014(−0.0753 to 0.0630)−0.0007(0.0547 to 0.0476)
  • ****p≤0.0001.

  • [Bold] represents significant beyond the 0.0001 credible interval. AIC, Aikaike information criterion; BIC, Bayesian information criterion; CAR, Conditional Autoregressive; Df, degrees of freedom.