Table 6

Cost comparison Expected number of undetected regions included in the true cluster E(s* - S), expected number of detected regions not in the true cluster E(L - S) and the ratio of costs C/C₂ (r = 1, 2) incurred by incomplete identification of the true cluster. The spatial scan statistic with low values is better.

Hot-spot Cluster

Scan statistic

E(s* - S)

E(L - S)

the ratio C/C₂

r = 1

r = 2

A = {14, 15, 20}

Flexible (K = 15)

0.108

2.951

3.059

3.167

Circular (K = 15)

0.065

0.722

0.787

0.852

B = {14, 15, 20, 26}

Flexible (K = 15)

0.097

2.548

2.645

2.742

Circular (K = 15)

0.735

2.525

3.260

3.995

C = {14, 15, 26, 27}

Flexible (K = 15)

0.492

2.243

2.735

3.227

Circular (K = 15)

1.736

3.153

4.889

6.625

D = {73, 74, 75, 76, 78}

Flexible (K = 15)

1.774

1.088

2.862

4.636

Circular (K = 15)

2.770

1.709

4.479

7.249

Tango and Takahashi International Journal of Health Geographics 2005 4:11   doi:10.1186/1476-072X-4-11