Skip to main content
. 2014 Aug 13;180(6):626–635. doi: 10.1093/aje/kwu180

Table 3.

Selected Ecometric and Spatial Properties of a Measure of Neighborhood Physical Disorder Computed From 1,826 Block Faces in 4 US Cities Using Google Street View Imagery Dating From 2007–2011

Location No. of Block Faces Audited Cronbach's α for Raw Measures Internal Consistency of IRT Model Median Level of Latent Disorder, z Scorea (IQRb) Moran's I Kriging Parameter
Cross-Validation Result
b (Nugget) Φ (Range) c (Sill –Nugget) Residual Mean (SD) Median Absolute Deviation
New York, NY 532 0.48 0.86 0.07 (–0.55 to 0.42) 0.041 0.19 5,000 0.14 −0.002 (0.54) 0.40
Philadelphia, PA 503 0.69 0.89 0.07 (–0.69 to 0.70) 0.092 0.35 10,000 0.30 −0.003 (0.66) 0.46
Detroit, MI 502 0.63 0.89 0.47 (–0.21 to 1.10) 0.063 0.18 3,000 0.52 0.001 (0.77) 0.53
San Jose, CA 289 0.48 0.87 −0.55 (–0.69 to 0.00) 0.030 0.09 1,000 0.27 −0.002 (0.57) 0.35
Overall 1,826 0.62 0.93 0.07 (–0.69 to 0.51) −0.002 (0.65) 0.45

Abbreviations: CA, California; IQR, interquartile range; IRT, item response theory; MI, Michigan; NY, New York; PA, Pennsylvania; SD, standard deviation.

a z scores were calculated from the posterior distribution of the disorder IRT model.

b 25th–75th percentiles.