Skip to content
How should I choose...
 
Notifications
Clear all

[Solved] How should I choose a probability distribution to represent uncertainty?

3 Posts
3 Users
6 Likes
1,489 Views
Avatar
Posts: 2
Customer
Topic starter
(@leroysmith)
Active Member
Joined: 3 years ago

Hi there, I am still learning Analytica and wondering which probability distribution I should pick for my model.  Is there documentation on recommended distributions for different applications?  Thanks for any help!

2 Replies
Avatar
Posts: 3
Admin
(@krista)
Member
Joined: 3 years ago

Take a look at Choosing a distribution page in the Analytica wiki.  For most purposes if the quantity is continuous, with or without bounds, it’s convenient to use the built-in function UncertainLMH(xlow, xmedian, xhigh) distribution. You specify 10th, 50th, and 90th percentiles for xlow, xmedian, and xhigh parameters. You can also specify plow as, for example 20%, in which case xlow and xhigh are the 20th and 80th percentiles. You can also set lower and/or upper bounds as lb and ub. It uses a simple but effective algorithm (from Hadlock & Bickel, or Tom Keelin) to create a smooth distribution between the specified values.  Hope this helps!  

Reply
Avatar
Posts: 8
Customer
(@agy79a6q5vx0nbjjxy0vsixsg9b3)
Active Member
Joined: 2 years ago

If you're talking about representing a subjective probability distribution, you probably can judge whether a crude first cut should be normal, lognormal, uniform, or triangular.  That's a lot better than just assuming normal distributions. One good possibility is sometimes what Krista Myers suggested.

Reply
Share: