Biology Reference
In-Depth Information
0.8
0.4
0.0
2007
2008
2008
2008
IV
II
III
IV
Time (weeks)
Figure 12.4
Quantiles of the predictive distribution. The dashed line indicates the 1-α/2 = 0.995 quantile
used for the surveillance in Figure 12.3. Triangles indicate the alarms.
(85,Inf)
(75,85)
(65,75)
(45,65)
(15,45)
(5,15)
(1,5)
(0,1)
2007
IV
2008
II
2008
III
2008
IV
Time (weeks)
Figure 12.5
Overview of aberration detection for all eight age group time series using the Farrington algo-
rithm with α = 0.01.
One way to simultaneously monitor all eight age groups is to monitor each
time series separately using, for example, the Farrington procedure. This is
done by the following code:
R> s.far.all <- farrington(momo, control = list(range =
phase2, alpha = 0.01, b = 5, w = 4))
A plot of the alarms for each time series provides a graphical overview as
shown in Figure 12.5. Monitoring each series independently as done above
ignores possible correlations of the time series. Furthermore, if one wanted
to keep the number of false alarms at the same level as for the surveillance of
a single series, one could, however, have used an α, being one-eighth of what
was used for the single time series case previously.
R> plot(s.far.all, type = alarm time, xlab = "time (weeks)")
 
Search WWH ::




Custom Search