Truncating the triangular distribution

Answered

Follow

Tony Zheng

• November 09, 2023 23:42

See the attached image below. When I use the 10%/90% for my triangular distribution input (input is in the vector form), the tail end fell into the negative values. This creates trouble when sampling the tail end of the distribution as my model does not allow for negative values. I only have the 10th/90th percentile input data so switching to 0%/100% is not going to work. 

Any good way to truncate the triangular distribution? I wrote an IF statement to force the sampled value back to zero in case of sampling a negative value. It works for now but I am looking for a better solution.

1

• Official comment

  

  Rick Kossik

  • November 10, 2023 01:10
  • Edited

  Well, by forcing it to 0 it is no longer triangular. It would actually be a bit odd looking (some fraction centered on 0).  But on the other hand, what is so special about a triangular?  You say you "only have the 10th/90th".  What does that mean? How were those values estimated?  Where do those numbers come from?  On what basis are you assuming the distribution is triangular (or "truncated triangular", which is a bit odd)?  What do you actually think the shape of the distribution should be?  Can you draw it?

  Comment actions Permalink

Please sign in to leave a comment.