Information Technology Reference
In-Depth Information
Most monitoring algorithms take the following form:
If the actual value exceeds some boundary value (the series mean, for
example, or the series mean plus one standard deviation),
And if the actual value exceeds the predicted value for three observation
periods in a row,
Sound the alarm (if the change like an epidemic is expected to be tem-
porary in nature) or recalibrate the model.
TO LEARN MORE
Almost always, a model developed on one set of data will fail to fit a
second independent sample nearly as well. Mielke et al. [1996] investi-
gated the effects of sample size, type of regression model, and noise-to-
signal ratio on the decrease or shrinkage in fit from the calibration to the
validation data set.
For more on leave-one-out validation see Michaelsen [1987], Weisberg
[1985], and Barnston and van den Dool [1993]. Camstra and Boomsma
[1992] and Shao and Tu [1995] review the application of resampling in
regression.
Watterson [1996] reviews the various measures of predictive accuracy.
