Using poisson distribution for arrivals ?

Question

question

Valu R asked Jun 21, '17 Valu R commented Jun 22, '17

Using poisson distribution for arrivals ?

Hi,

I'd like to represent an arrival scheme for patient by day of week and time. I want to represent a model for a one month period.

To include variation in the patients arrivals I chose the custom arrivals from FlexSim and included a mean for each PCI and time interval. (I calculated this mean based on a one year data)

In the attached printscreen, please find my arrival scheme. This method doesn't seem to be effective because patients keep arriving at the same pace and same volume when I run different times the model on a one month period (using repeat interval weekly : 10080)

Could you please give me recommandations to calculate these number of arrivals by PCI and arrival time for custom arrivals on FlexSim and which distribution I should choose : poisson for emergency and exponential for elective ?

Thanks in advance for your support,

Valentin

Software Version:

FlexSim HC 5.0.12

arrival patterns poisson distribution

printscreen-custom-arrivals.png (235.9 KiB)

· 2

Matthew Gillespie ♦♦ commented · Jun 21, 2017 at 05:21 PM

Are you saying that each week has exactly the same arrivals? OR are you saying that each week is different, but that every time you run the model you get the same results?

0 ·

Valu R commented · Jun 21, 2017 at 07:34 PM

@Matthew Gillespie Thanks for your reply. Each week is different but every time I run the model I do get the same results. Beside, arrival times for a specific week do not change when running several time the model. For e.g. Week 1 a patient with PCI 1 arrives always Monday at 11.30 a.m. Arrival times between weeks (week 1, week 2 etc.) are however different.

0 ·

Answer 1 · 2017-06-21T19:41:16Z

Matthew Gillespie answered Jun 21, '17 Valu R commented Jun 22, '17

That is by design. By default random streams repeat every model run so that your model is repeatable. This is very beneficial when building the model since you can debug the model and when you make changes to the model you can see the effects of those changes.

When you have your model to a state where you want to begin using different random numbers to gather data you should run an experiment. Running an experiment lets you run many replications on different scenarios of your model. These replications also repeat random streams so that you can debug any issues with a particular replication.

If you really need to you can disable the repetition of random streams in the Edit menu, however, for most cases you should just run an experiment with more replications to gather data since it is impossible to debug any issues you run into while running an unrepeatable random stream.

· 1

Valu R commented · Jun 22, 2017 at 09:11 AM

@Matthew Gillepsie Thanks a lot for your helpful pieces of advice and information. I now better understand how and why the arrival patterns work this way on FlexSim.

0 ·

question