LAB 05 / GAMMA-POISSON UPDATE
Event counts need observation time
Ten events in one hour versus ten hours: what event rate per hour does each suggest?
← Seven labs / LAB 05About estimating an event rate
Read counts together with exposure time
The target is the average number of events per hour. A gamma–Poisson update combines the observed count and exposure time with the prior to update the distribution over event rates.
Hold the count fixed and change the observation time
Density near zero can exceed the vertical plotting limit, so it is clipped. Do not use the plot to compare areas under the curves.
Prior distribution
Observed event rate (vertical line)
Posterior distribution
Observed event rate-
Posterior mean-
95% credible interval-
Does zero observed events imply a zero event rate?
With zero observed events, a longer observation time supports lower event rates more strongly and narrows the yellow curve. Zero events alone do not establish that the rate is zero.