Detecting changes in incidence rate

[Gerenuk]

Hi!
I have rare events over time (about 200 incidences) and I'd like to detect when the time rate of these increase, i.e. at which point more events per time start to occur.
Do you know a solid statistical tool to analyze that? I'd also like to know if the change is real evidence or just random.

 

[SW VandeCarr]

Hi!
I have rare events over time (about 200 incidences) and I'd like to detect when the time rate of these increase, i.e. at which point more events per time start to occur.
Do you know a solid statistical tool to analyze that? I'd also like to know if the change is real evidence or just random.

It's somewhat dependent on the temporal profile: isolated random spikes, periodic spikes or waves, or sustained trends. Generally, various applications of moving averages are often used. These are fairly robust and can be adapted to different profiles. You can look up "Moving Average" on the Wiki for a fairly good general discussion. Linear regression models also work well for more sustained trends where you want to test for trends greater or less than 0.

The following paper shows an application of a combined approach called Autoregressive Integrated Moving Average (ARIMA) for dengue fever.

http://www.ajtmh.org/content/79/6/933.full.pdf

EDIT: As an aside, sometimes just inspection is more appropriate than statistical testing if public safety is an issue. You really want to minimize beta error (missing a trend that actually exists). You also will need to control for confounding factors. Differences in reporting and/or detection over time is often the case.

 

[Gerenuk]

Thanks a lot! I hope I can make use of the information.

My data is most likely something like a constant rate up to some time and then an increase in frequency. The problem is that it is only about 200 events in total.

For regression I'm afraid it won't tell me if the trend is with the confidence interval. But I need to read more about it...

 

[awkward]

You also might consider a control chart.

http://en.wikipedia.org/wiki/Control_chart

 

[blue_raver22]

Hello there,

Thank you for reaching out with your question. I would recommend using a statistical tool called a Poisson regression model to analyze your data. This type of model is commonly used for rare events and can help identify changes in incidence rates over time.

To determine if the change in incidence rate is significant, you can also perform a hypothesis test using the model results. This will help determine if the change is likely due to random chance or if there is a real underlying trend.

It is also important to consider potential confounding factors that may be influencing the incidence rates, such as changes in population size or demographics. These should be taken into account in your analysis to ensure accurate results.

I hope this helps and good luck with your research!