Friday, June 7, 2019

The core of knowledge

... statistics is what tells you if something is true, false, or merely anecdotal; ... it is the instrument of risk-taking
Statistical and applied probabilistic knowledge is the core of knowledge;
Nassim Taleb

You got to hand to the guy: he is passionate about his subject.

Of course, he's also the first person to tell you that his most infamous invention, the "Black Swan", is black - - that is, extraordinarily rare - - because there are no statistics to predict it.

Statistics in the PMO
Two of the more useful properties of statistics are these:
  • They are are surrogates of a past track record; thus, handy for estimating
  • They are persistent in the sense that they are survivors: wait weeks or months and if the circumstances haven't changed, neither have the statistics. Thus, your story doesn't change with time, an idea called "stationary".
Rules of thumb
For project purposes, rules of thumb may be as useful as actual statistical analysis:
  • There's shift-right bias at milestones; milestones are inherently weak objects in the schedule
  • Diversification follows the "square-root-of-N" rule: diversify into 4 independent events, and get a 2:1 improvement in risk (square root of 4 is 2)
  • "Expected value" -- which is a risk weighted average -- is usually more pessimistic, and therefore more conservative, than "most likely" value" (the single value with highest individual probability)
  • "Expected utility value" is not objective, and subject to many biases (utility is a weighting according to usefulness)
  • Independent events tend to centrally cluster--the theory of central tendency
  • A sample of a large population can be an economic approach that saves time and money

Buy them at any online book retailer!