Can somebody help estimating how many analysis units (number of railway level crossings - RLCs) will I need to build a crash prediction model (safety performance function)? From some past analyses of few RLCs they seem to have on average 0.03 crash/RLC/year. (Note: I have 9 available years of crash data).
Thank you, Waldemar. In the first round, we aim for 100 RLCs - I wonder if they will suffice :-) In similar studies which I saw, they usually had thousands of RLCs...
The chance of occurring an event is very rare ~ 0.03 (nearly zero). For rare events , applying the method of general sample size calculation is bit insufficient as it does not account for probability nearly zero for a bernoulli event and very small variance.
For rare events , you definitely need a large sample size(which may be you found in other studies). There are online calculators available to calculate the sample size for rare events for reference. A higher cost is always involved in getting large sample. You may need to work out what suits you best in terms of statistical accuracy , power and budget.
Dear Aarti, thank you very much. Can you please provide a link to some calculator, as you mention? Thanks!
https://www.statstodo.com/SSizRareInc_Pgm.php
Try this link. I hope it will work and help you out.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size