[kplato] definitions of effort and risk

[Jim Sabatke]

John,

You are quite correct.  For this application risk is a beta 
distribution.  I have used software that has a std. dev. column and it 
is so counter-intuitive for most people that I think it is useless.

I totally agree with your analysis.

Jim

John D Lamb wrote:

>Thanks. I think I've also worked out what risk is:
>
>Risk: a simple estimate of the shape of the distribution, given
>optimistic, pessimistic and expected efforts.
>
>This fits with my earlier thought that "risk" could be represented by a
>Distribution superclass (at least from the "document" or
>"implementation" point of view: "Risk" is probably better from the users
>point of view). Zero risk would be a fixed value, low risk is probably
>beta-shaped and high risk lognormalish. The precise shape of the
>distribution would not be important. From the implementation point of
>view, it might be useful to represent risk by some real distribution
>that gave the same values for duration, but could also give random
>values and confidence intervals if later we need them: i.e. "fit
>E(p,o,e) to a curve and calculate the risk from that".
>
>I would definitely try to define risk without using standard deviations.
>Standard deviations are good for *measuring* uncertainty, but not
>intuitive for most people when *specifying* uncertainty. Optimistic,
>Pessimistic, expected and a "Risk" factor ought to work just fine.
>
>We might have to use Monte-Carlo simulation for a "project confidence
>interval". But I think it would be sensible to get an initial project
>first.
>
>JDL
>
>Jim Sabatke wrote:
>
>>Actually, because E(p) is usually a good bit larger than E(o), the
>>distribution is skewed.
>>Also, come to think of it, we should be calculating a confidence value
>>so that we can multiply them up the chain to obtain an ovall project
>>confidence.
>>
>>Jim
>>
>>bilbo wrote:
>>
>>>On 21 Jun 2001, at 20:15, Jim Sabatke wrote:
>>>
>>>>I guess the tabs didn't expand well for you.  Sorry.
>>>>
>>>My problem, probably, but I try to avoid tabs :-)
>>>
>>>>Most PM software expects the user to enter either:
>>>>
>>>>1.    E(p), E(e) and E(o)
>>>>2.    E(p), E(o) and a std. dev. value, which is way above most PM's
>>>>(IMHO) to use correctly.
>>>>
>>>Isn't that assuming symmetrical distribution of E(p) and E(o)?
>>>
>>>>You are correct that the values would be entered based on experience.  I
>>>>am actually trying to create something more intuitive to the average PM
>>>>by adding a risk factor column.  The PM wouldn't need the math, just a
>>>>comparative judgement call (high/low/normal risk).
>>>>
>>>The sw person in me wants to try to fit E(p,o,e) to a curve and
>>>calculate the risk from that. I don't think that's going to work.
>>>
>>>Thanks for explaining that.
>>>regards,
>>>       Bill
>>>
>_______________________________________________
>kplato mailing list
>kplato@master.kde.org
>http://master.kde.org/mailman/listinfo/kplato
>
>

-- 
Jim Sabatke
SuSE 7.1 Linux
Kernel - 2.4.0
http://www.execpc.com/~jsabatke

"First God made the idiot; that was for practice. Then he made school boards." Mark Twain