Tips on using this forum..

(1) Explain your problem, don't simply post "This isn't working". What were you doing when you faced the problem? What have you tried to resolve - did you look for a solution using "Search" ? Has it happened just once or several times?

(2) It's also good to get feedback when a solution is found, return to the original post to explain how it was resolved so that more people can also use the results.

Best Duration Uncertainty to use – Triangle or Trigen

7 replies [Last post]
claire ingram
User offline. Last seen 5 years 37 weeks ago. Offline
Joined: 24 Jan 2011
Posts: 2
Groups: None

I am wondering if anyone can help me please?  I am working on a Wind Farm project and putting the schedule through Primavera Risk analysis .  What is the best Duration Uncertainty to use – Triangle or Trigen? Thank you Claire

Replies

Dennis Hanks
User offline. Last seen 3 years 30 weeks ago. Offline
Joined: 17 Apr 2007
Posts: 310

You have my sympathy.

Stephen Devaux
User offline. Last seen 5 weeks 1 day ago. Offline
Joined: 23 Mar 2005
Posts: 624

Huh. That there's pretty rude, in't?

Dunning-Kruger is a sad thing, and pretty much incurable.

Cheeri-O.

Steve the Bajan

Dennis Hanks
User offline. Last seen 3 years 30 weeks ago. Offline
Joined: 17 Apr 2007
Posts: 310
Steve; Where to begin. Dennis wrote: "...I would use the trigen if (big if) I thought the ranges given were understating the risk - my respondents were risk accepting, giving optimistic estimates.  I would use triangle if I thought they fully understood/appreicated the risk(s)." Personal approach to risk varies - some are timid, others reckess. Part of my role is to moderate extremes.That's sort of the big problem with Monte Carlo risk simulation packages, isn't it? Unlike you, most of us don't know the future. We have to guess. I like to think we make informed quesses based on experience, but they're still guesses.  Nobody knows what the objectively "correct" number is, so if we run it on one default and don't like the answer, we tweak it until we get the answer we wanted all along. In other words, we decide what number we want and then use pseudo-science to justify our gut instincts. it's the classic example of "garbage in, gospel out". Maybe in your world, not mine. They may not like the answer, but they are unlikely to manipulate it. I expect them to bounce the result against their gut. I have a healthy respect for years of experience and practiced judgement. If the result doesn't 'smell right', I'd look at our assumptions. Review is part of the process. Pseudo-science?Monte Carlo systems can provide valuable information if:There is good historical data showing activity durations and effort on similar work, and: The function of the estimate.We map the distribution curve to each specific activity. PRA does this easily, not sure what software you're using.They either include no lags or, as with Spider Project, the lags can be volume lags that can be varied according to the simulated distribution. Post-FEED schedules should have few lags. Pre-FEED plans have few activities - lags are managable. PRA seems to handle lags well. But almost no one does any of that -- it’s too much work. Instead, they pluck three estimates out of thin air and then they run the simulation using a default. Really? Out of thin air? We don't ask people off the street. Respondents have knowledge, experience, judgement. And if they don't like the results fromthat default, they use another one. Not my experience. And, as Dennis suggests, the whole purpose is to satisfy a customer who knows nothing about the process but is impressed when told: "It must be accurate -- it came out of a computah!"  Don't think I ever suggested this. My clients are more sophisticated than this, are you projecting? I hesitate to suggest this, given your penchant for fabrication, but if you have twenty minutes this YouTube video might explain my approach to post-FEED schedule risk assessment - Contigency. https://www.youtube.com/watch?v=amnxmkl1c3c Dennis
Stephen Devaux
User offline. Last seen 5 weeks 1 day ago. Offline
Joined: 23 Mar 2005
Posts: 624

Dennis wrote: "...I would use the trigen if (big if) I thought the ranges given were understating the risk - my respondents were risk accepting, giving optimistic estimates.  I would use triangle if I thought they fully understood/appreicated the risk(s)."

That's sort of the big problem with Monte Carlo risk simulation packages, isn't it? Nobody knows what the objectively "correct" number is, so if we run it on one default and don't like the answer, we tweak it until we get the answer we wanted all along. In other words, we decide what number we want and then use pseudo-science to justify our gut instincts. it's the classic example of "garbage in, gospel out".

Monte Carlo systems can provide valuable information if:

  1. There is good historical data showing activity durations and effort on similar work, and:
  2. We map the distribution curve to each specific activity.
  3. They either include no lags or, as with Spider Project, the lags can be volume lags that can be varied according to the simulated distribution.

But almost no one does any of that -- it’s too much work. Instead, they pluck three estimates out of thin air and then they run the simulation using a default, And if they don't like the results fromthat default, they use another one. And, as Dennis suggests, the whole purpose is to satisfy a customer who knows nothing about the process but is impressed when told: "It must be accurate -- it came out of a computah!"

Fraternally in project management,

Steve the Bajan

Waleed Mahfouz
User offline. Last seen 3 years 23 weeks ago. Offline
Joined: 8 Sep 2007
Posts: 28
Groups: None

The best PDF to represent activities durations is Lognormal distribution

Please refer to the following paper

Construction Activities Duration Patterns in the Middle East http://www.sapub.org/global/showpaperpdf.aspx?doi=10.5923/j.ijcem.20130204.04
Dennis Hanks
User offline. Last seen 3 years 30 weeks ago. Offline
Joined: 17 Apr 2007
Posts: 310

Claire:

For me, the difference is whether or not my respondents are risk adverse, risk accepting, or neutral.  Of the two possible distrubutions you cite, I would use the trigen if (big if) I thought the ranges given were understating the risk - my respondents were risk accepting, giving optimistic estimates.  I would use triangle if I thought they fully understood/appreicated the risk(s).  Since it is relatively easy to change the risk distributions in the layout, I might run both just for grins. I would run BetaPert to model the risk adverse case.

Hope this helps.

Rafael Davila
User offline. Last seen 5 hours 22 min ago. Offline
Joined: 1 Mar 2004
Posts: 4792