Bureau of Transportation Statistics (BTS)
BTS Home > Publications > Journal of Transportation and Statistics > Volume 4 Number 1 > Paper 3
Printable Version

Table 2 - Input Parameters for 38 Trial Cluster Analyses on Land-Use Distress

Excel | CSV

Case Distance Linage Standardization No. of clusters Comments
1 Euclidean Single None 20 Produced single large clusters of over 300 districts with no cluster larger than 5 districts; chaining effect took place
2 Euclidean Single None 15 Produced single large clusters of over 300 districts with no cluster larger than 5 districts; chaining effect took place
3 Euclidean Single None 25 Produced single large clusters of over 300 districts with no cluster larger than 5 districts; chaining effect took place
4 Euclidean Average None 20 Smaller main clusters and larger minor clusters but not much diversity in residential scenarios
5 Euclidean Average None 30 Smaller main clusters and larger minor clusters but not much diversity in residential scenarios
6 Euclidean Average None 18 Smaller main clusters and larger minor clusters but not much diversity in residential scenarios
7 Euclidean Centroid None 20 Relatively poor distribution of cluster sizes and poor diversity in residential scenarios
8 Euclidean Centroid None 30 Relatively poor distribution of cluster sizes and poor diversity in residential scenarios
9 Euclidean Complete None 20 Somewhat more diversity in residential scenarios
10 Euclidean Complete None 17 Somewhat more diversity in residential scenarios
11 Euclidean McQuitty None 20 Little diversity in residential scenarios
12 Euclidean Median None 20 Large mega-cluster, like with single linkage
13 Euclidean Ward None 20 Relatively equal cluster sizes; good diversity
14 Pearson Average None 20 Pearson and Squared Pearson linkages created mega-clusters; Manhattan seemed almost as good as Euclidean; Squared Euclidean was nearly identical to Euclidean
15 Manhattan Average None 20 Pearson and Squared Pearson linkages created mega-clusters; Manhattan seemed almost as good as Euclidean; Squared Euclidean was nearly identical to Euclidean
16 Sq. Euclidean Average None 20 Pearson and Squared Pearson linkages created mega-clusters; Manhattan seemed almost as good as Euclidean; Squared Euclidean was nearly identical to Euclidean
17 Sq. Pearson Average None 20 Pearson and Squared Pearson linkages created mega-clusters; Manhattan seemed almost as good as Euclidean; Squared Euclidean was nearly identical to Euclidean
18 Pearson Complete None 20 Pearson and Squared Pearson linkages created identical mega-clusters; Manhattan created slightly more diversity than Euclidean; Squared Euclidean identical with Euclidean
19 Manhattan Complete None 20 Pearson and Squared Pearson linkages created identical mega-clusters; Manhattan created slightly more diversity than Euclidean; Squared Euclidean identical with Euclidean
20 Sq. Euclidean Complete None 20 Pearson and Squared Pearson linkages created identical mega-clusters; Manhattan created slightly more diversity than Euclidean; Squared Euclidean identical with Euclidean
21 Sq. Pearson Complete None 20 Pearson and Squared Pearson linkages created identical mega-clusters; Manhattan created slightly more diversity than Euclidean; Squared Euclidean identical with Euclidean
22 Pearson McQuitty None 20 Changing the distance measure has similar effects as with average and complete linkages
23 Manhattan McQuitty None 20 Changing the distance measure has similar effects as with average and complete linkages
24 Sq. Euclidean McQuitty None 20 Changing the distance measure has similar effects as with average and complete linkages
25 Sq. Pearson McQuitty None 20 Changing the distance measure has similar effects as with average and complete linkages
26 Pearson Ward None 20 Pearson and Squared Pearson produced slightly less diversity in residential and agricultural scenarios; Squared Euclidean lumped agricultural scenarios together; Manhattan similar to Euclidean
27 Manhattan Ward None 20 Pearson and Squared Pearson produced slightly less diversity in residential and agricultural scenarios; Squared Euclidean lumped agricultural scenarios together; Manhattan similar to Euclidean
28 Sq. Euclidean Ward None 20 Pearson and Squared Pearson produced slightly less diversity in residential and agricultural scenarios; Squared Euclidean lumped agricultural scenarios together; Manhattan similar to Euclidean
29 Sq. Pearson Ward None 20 Pearson and Squared Pearson produced slightly less diversity in residential and agricultural scenarios; Squared Euclidean lumped agricultural scenarios together; Manhattan similar to Euclidean
30 Sq. Euclidian Ward Z-scores 20 Similar diversity to case without standardization (case 28) but oddly distributed variables better represented
31 Euclidean Complete Z-scores 20 Much poorer diversity than in case 9
32 Euclidean Average Z-scores 20 Forms mega-cluster; worse than case 4
33 Euclidean Ward Z-scores 20 More diverse in some areas than with case 13
34 Sq. Euclidean Ward Z-scores 30 Improved diversity over case 30
35 Euclidean Ward Z-scores 30 More diverse than case 33
36 Euclidean Ward None 30 Similar diversity to case 35 but oddly distributed variables like R3, R4 not as well represented
37 Euclidean Ward Scaled percentages 30 Oddly distributed variables well-represented but not enough of an improvement in variable bounds
38 Sq. Euclidean Ward Z-scores 35 Number of clusters increase to 35 to separate a few odd groupings
Research and Innovative Technology Administration (RITA) • U.S. Department of Transportation (US DOT)
1200 New Jersey Avenue, SE • Washington, DC 20590 • 800-853-1351 • E-mail RITA

Accessibility | Disclaimer | Fast Lane | FedStats | Freedom of Information Act | No FEAR Act | OIG Hotline | Privacy Policy | RSS | USA.gov | White House | Wireless

Plug-ins: PDF Reader | Flash Player | Excel Viewer | PowerPoint Viewer | Word Viewer | WinZip