TY - GEN
T1 - Testing differences between case and control point patterns using nearest neighbour distances and bootstrapping
AU - Henry, KA
AU - Burge, LM
AU - Nguyen, D
PY - 2003
Y1 - 2003
N2 - This paper proposes a method for comparing point patterns in case and control data by using nearest neighbour distances, bootstrapping and the Wilcoxon Signed-Rank Test. Case-control comparisons are frequently used in medical geography and epidemiology to examine the patterns of disease and infer transmission of infectious pathogens. Our strategy addresses the problem of handling spatial analysis when the numbers of cases differ from controls. Differences in sample sizes can affect density of points and therefore bias nearest neighbour distances. To demonstrate this method we created a control set of 250 points and two sets of cases 125 points each. Bootstrapping the control data and comparing each run statistically to the cases can provide confidence intervals and estimate the risk of erroneously rejecting the null hypothesis. We follow with a case study of tuberculosis. The spatial distributions of different bacterial strains were compared and the nearest neighbour distances were analyzed as a surrogate for possible transmission of tuberculosis. The method may be useful to epidemiologists, geologists, biologists, geographers and ecologists for evaluating differences between the spatial structures of points.
AB - This paper proposes a method for comparing point patterns in case and control data by using nearest neighbour distances, bootstrapping and the Wilcoxon Signed-Rank Test. Case-control comparisons are frequently used in medical geography and epidemiology to examine the patterns of disease and infer transmission of infectious pathogens. Our strategy addresses the problem of handling spatial analysis when the numbers of cases differ from controls. Differences in sample sizes can affect density of points and therefore bias nearest neighbour distances. To demonstrate this method we created a control set of 250 points and two sets of cases 125 points each. Bootstrapping the control data and comparing each run statistically to the cases can provide confidence intervals and estimate the risk of erroneously rejecting the null hypothesis. We follow with a case study of tuberculosis. The spatial distributions of different bacterial strains were compared and the nearest neighbour distances were analyzed as a surrogate for possible transmission of tuberculosis. The method may be useful to epidemiologists, geologists, biologists, geographers and ecologists for evaluating differences between the spatial structures of points.
UR - https://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=purepublist2023&SrcAuth=WosAPI&KeyUT=WOS:000184327900004&DestLinkType=FullRecord&DestApp=WOS
UR - http://www.scopus.com/inward/record.url?scp=35248869323&partnerID=8YFLogxK
U2 - 10.1007/3-540-44842-x_4
DO - 10.1007/3-540-44842-x_4
M3 - Conference contribution
SN - 3540401563
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 33
EP - 42
BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
A2 - Kumar, Vipin
A2 - Gavrilova, Marina L.
A2 - Kenneth Tan, Chih Jeng
A2 - L’Ecuyer, Pierre
A2 - Kenneth Tan, Chih Jeng
PB - Springer Verlag
ER -