Skip to main content

Table 1 Geographic profiling and other measures of spatial central tendency

From: Geographic profiling as a novel spatial tool for targeting infectious disease control

 

Geographic profiling

Other methods of spatial central tendency

Site

Hit score

SSA (Spatial mean)

SSA (Spatial median)

SSA (Centre of minimum distance)

(a) London

    

Broad Street pump

0.20%

5.10%

7.40%

5.20%

(b) Cairo

    

Source 1

2.80%

4.50%

4.00%

4.40%

Source 2

1.30%

5.87%

5.35%

5.85%

Source 3

2.18%

0.42%

0.11%

0.38%

Source 4

2.53%

5.17%

4.68%

5.14%

Source 5

3.06%

5.22%

4.72%

5.19%

Source 6

3.06%

5.16%

4.67%

5.13%

  1. A comparison of the geographic profiling model's performance against measures of spatial central tendency for (a) the London cholera data and (b) the Cairo malaria data. For each source of infection (the Broad Street pump in the London analysis, and the seven water sources that tested positive for An. sergentii in the Cairo analysis), the table shows the model's hit score percentage in comparison to the spiral search area (SSA) percentage for three measures of spatial central tendency, with the most efficient search marked in red. The SSA is obtained by calculating the percentage of the total area under consideration that has to be searched before locating the correct source, moving out equally in all directions from, respectively, the spatial mean, spatial median and centre of minimum distance; it is thus equivalent to hit score percentage in geographic profiling (see Methods). In the London analysis, GP located the correct source after searching a smaller percentage of the surrounding area (0.2%, compared to 5.1%, 7.4% and 5.2% for the other three methods respectively). In the Cairo analysis, geographic profiling performed better than all three other methods for five out of seven sources, and was almost the best for one of the other two sources (4.46% versus 4.41%). Only for one source, which happened to lie close to the centre of the disease case locations, did the model not perform as well.