Biology Reference
In-Depth Information
2.5.2.1 Single-Day Spike ............................................................... 35
2.5.2.2 Lognormal Outbreak Signature...................................... 35
2.5.2.3 Adding Effects to Initial Outbreak Signatures ............. 35
2.6 Example: Mimicking a BioALIRT Dataset ............................................... 36
2.6.1 Mimicking Background Health Data ........................................... 36
2.6.2 Distribution Testing......................................................................... 39
2.6.3 Outbreak Insertion .......................................................................... 39
2.7 Summary and Future Work ....................................................................... 47
2.7.1 Future Work...................................................................................... 47
2.7.2 Summary........................................................................................... 48
Acknowledgments ................................................................................................ 48
References............................................................................................................... 49
KEYWoRDS
Simulation, Time series, Multivariate, Goodness of fit,
Disease outbreak
2.1 Motivation
The field of biosurveillance involves the monitoring measures of diagnos-
tic and prediagnostic activity for the purpose of finding early indications
of disease outbreaks. By providing early notification of potential outbreaks,
the aim is to provide public health officials the opportunity to respond ear-
lier and thus more effectively. Although the field has grown in importance
and emphasis in the past several years, the research community involved in
designing and evaluating monitoring algorithms has not grown as expected.
A major barrier has been data accessibility: typically researchers do not have
access to biosurveillance data unless they are part of a biosurveillance group.
In fact, after parting from a biosurveillance group, researchers lose their data
access. This means that a very limited community of academic researchers
works in the field, with a nearly impenetrable barrier to entering it (espe-
cially for statisticians or other nonmedical academics). Furthermore, the con-
finement of each research group to a single source of data and the lack of
data sharing across groups “leaves opportunity for scientific confounding”
(Rolka, 2006).
While simulated data have their own difficulties, they seem to be a neces-
sity for modern biosurveillance research. Buckeridge et al. (2005) explain:
[They] are appealing for algorithm evaluation because they allow exact
specification of the outbreak signal, perfect knowledge of the outbreak
onset, and evaluators can create large amounts of test data …
