Activity Number:
|
279
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Statistics in Epidemiology*
|
Abstract - #301801 |
Title:
|
Methods for Computing Exact P-Values of the Range-Scan Statistic
|
Author(s):
|
James Mancuso*+
|
Affiliation(s):
|
Pfizer Global R&D
|
Address:
|
Eastern Point Rd, MS 8260-2202, Groton, Connecticut, 06340, USA
|
Keywords:
|
Range-scan test ; Adverse events ; Time series ; Cluster ; Exact test
|
Abstract:
|
Range-scan statistics are useful for the detection of temporal clustering of infrequent events such as adverse events in a clinical trial or the incidence of a rare disease in a particular community. The value of the statistic is the length of the shortest window of time containing all events. Exact methods have been proposed for the computation of P-values under two discrete incidence models. The occupancy model assumes a stable population at risk (PAR) for the duration of study while the structured environment involves a PAR that changes irregularly over time. This paper derives a new computational formula for the exact P-value of the range-scan statistic in the structured environment model. The methods are implemented in the C++ programming language, thereby extending their practical applicability to larger data sets. Applications and generalizations of the methodology are discussed.
|