My team collected data at Prarie Fire in Union South from 8-9 AM on Tuesdays and Thursdays. We noticed that more people would arrive towards the end of our shift than in the beginning. When all of these are pooled together, it resembles a beta distribution, but this does not take into account the fact that short interarrival times are clustered at the back end of the time. When we run our model, lines do not build up like we saw because of the shift. Does anyone have any suggestions on how to handle this situation?
Another way to deal with the situation would be to simply split up the data into two different models. If your goal is simply to help Prairie Fire improve their process, then using two models may more accurately reflect how the system is working, and help Prairie Fire reallocate their workers and resources accordingly.
This is a really old post but I was curious if anyone could explain further on how would you split this data to create two models? Are you suggesting just having two models one for the first part of the shift and one for the end