|Saturday, February 22|
|CS18 Modeling Events in Time||
Sat, Feb 22, 9:00 AM - 10:30 AM
Is event interval analysis history? (302715)*James Arthur Lemon, University of NSW
Keywords: uncommon events, rate analysis, GLM
Investigating rates of events using the Poisson distribution to model counts of the events of interest in equally spaced time intervals is a standard procedure. Where events are independent, numerous, and have a constant rate, this technique is generally adequate. Problems may arise when the events of interest are uncommon, leading to small counts within intervals.
In the middle of the twentieth century, there was considerable interest in analyzing the intervals between events when modelling rates, known then as event history analysis. Given the great improvements in computational algorithms since then, it may be useful to consider this technique when time-varying rates and data segmentation combine to make analysis by counts less reliable.
Comparative analyses of uncommon events such as hurricanes demonstrates that event interval analysis may reveal rate changes with fewer events, and that smaller changes in rates can be revealed.