13 November 2009

Precision Is Frivolous

It struck me the other day that extreme precision is frivolous.

Engineers in particular tend to overvalue precision. I have three engineering degrees, so I think I can speak with some authority here. And I can attest to the fact that when engineers (and others) use extremely precise data, they are often overstating the value of their analysis. Just because they measured the distance between A and B to the micrometer doesn't mean they measured it in a meaningful way.

In mathematics, we talk about "significant figures," but we probably are not sufficiently aware of the presence of "insignificant figures." That is, 3.1415926 is pi to 7 "significant" figures... but depending on the application, those last 5 digits may or may not actually be significant.

So don't tell me it's exactly a 3-minute walk from here to there, when you know full well it takes "approximately 5 minutes." I don't want to hear that the wind is blowing at 42 knots, when clearly it's gusting to roughly 40. And really, finishing 33% of a project isn't any different at all from finishing 30% or 35%.

Don't get me wrong - I like frivolity when it's fun, and I like precision when it's necessary. But I have no interest in the grim frivolity of unnecessary precision.


Glen B. Alleman said...


DID 81650 mandates a statistical risk analysis that forces the conversation about cost, schedule, and technical performance to be

"We have an 82% confidence of completly on or before 13 Nov 2009"

Mark said...

"Significance" is a funny word.

In terms of statistical hypothesis testing, something I work with and coach colleagues on everday, the question is usually of the form "is there a significant difference"? Too often people confuse statistically significant with practically significant. They might get a result that indicates yes, there is a difference, and celebrate (assuming a difference is good). But if that difference is too tiny to have any practical impact, all they are really celebrating is that they were able to see a very tiny difference. Congratulations. Big deal.

The Dan Ward said...

@Glen - Good point... and I'm just sayin' that's a frivolous degree of specificity, and not in the good way. :)

@Mark - Great example! Man, I would love to shadow you at work for a day and see you in action.

Gabe said...

Funny phrase, "grim frivolity of unnecessary precision."

It makes me smile!

Sean Walsh said...

Your entry reminded me of Materials Lab Senior Year of engineering school. It was Fall 1975 and we all had relatively new TI SR-50 calculators whearas before we had been using engineering slide rules. The experiment was to measure the Young's Modulus of a 2X4. We all took the measurements and calculated the results and copied the results from our calculators into our lab reports. Our lab reports all came back with lots of red and a lecture from the instructor about significant figures. I realized then that one of the advantages of slide rules was that that they weren''t as precise.

Another lesson was in grad school when we had to take an underwater acoustics course as part of our naval ship design program. Most of us had an ME background and the professor scoffed at us when we started pulling out calculators to work problems. For the sonar equation and ocean acoustic environment, doing approximate conversions to dB and adding and subtracting dB was close enough.