||5R01CA095979-05 Interpret this number
||Harvard Pilgrim Health Care, Inc.
||Tests for Spatial Randomness in Cancer Maps
DESCRIPTION (provided by applicant): Cancer maps can provide important clues
concerning geographical variability in the etiology, prevention, screelling or
treatment of cancer. As in all medical research, it ts important to determine
whether any variation observed may reasonably be due to chance or not. This can
be done using tests for spatial randomness, adjusting for the uneven
geographical population density. Many such tests have been proposed, but for
most, little is known about their properties, and they are seldomly used in
cancer atlases. In this methodological project we will (i) develop theoretical
properties that any test for spatial randomness should fulfill in order to be
useful for cancer maps, (ii) determine which test statistics do and do not
fulfifi these properties, (iii) evaluate the statistical power of different
test statistics for different alternative hypotheses, (iv) determine the
ability of different tests to estimate cluster model parameters when the null
hypothesis is rejected, and (v) evaluate and ifiustrate the practical use of
different test statistics on, among other data sets, county based brain cancer
mortality data from the United States and individually geocoded breast cancer
treatment data from Connecticut.
Cancer map patterns: are they random or not?
Kulldorff M, Song C, Gregorio D, Samociuk H, DeChello L
Am J Prev Med, 2006 Feb;30(2 Suppl), p. S37-49.
Brain cancer mortality in the United States, 1986 to 1995: a geographic analysis.
Fang Z, Kulldorff M, Gregorio DI
Neuro Oncol, 2004 Jul;6(3), p. 179-87.
NCI NIH HHS - R01-CA95979-0