Tuesday, October 21, 2008

Eyeballing

In a comment, Mr. Idiot points us all at this profound illustration on the accuracy of "eyeballing" measurements.

Your inaccuracy by category:





Parallelogram------------
Midpoint------------
Bisect angle------------
Triangle center------------
Circle center------------
Right angle------------
Convergence------------

Average error: ---- (lower is better)


10 comments:

GMR said...

The eyeballing tests are simplty effective at keeping one's ego in check! My first total score was 16.27. Ouch! Then a significant improvement to 7.21 immediately following the first trial. After several hours at the end of the day did I really learn or retain information -- I scored a 6.50. LNDD- AFLD its your turn for disclosure.

Thomas A. Fine said...

Hmmm.

I started in the 3's somewhere. And got better from there. I improved a lot in most categories with practice. Best score was 1.9 something.

Just for fun, I tried cheating (using a piece of paper to line things up). My score definitely improved, but not a lot.

I'm just a freak of nature, I guess. Maybe I should go work for the labs...

Cycling Fan said...

I also got in the low 2's but I think this disqualifies me from the labs.

Unknown said...

Would have been an interesting device to put LNDD witnesses (those that supported eyeballing or used their experience) to the test in Malibu........

Unknown said...

7.43 Couldn't get the parallelogram (however you spell it...

Mike Solberg said...

One of the things that interests me about the game is the standard deviation, at least for me. I can really nail it once in a while, but then the next shape will be way off. I've done it enough to realize it has little to do with "experience." I'm just inconsistent, given the nature of the game. You know, I don't even remember what EXACTLY our friends in the lab at LNDD were "eyeballing," but I don't remember ever seeing any evidence of their eyeballing accuracy.

syi

Laura Challoner, DVM said...

I call this game Mongongu, like Soduku......... only different.

Mike Solberg said...

Nice Bill, that's funny!

syi

Cub said...

The eyeballing game shows you how far off you are, and that feedback at least gives you the possibility of improving your eyeballing skills.

I don't think you can say the same for eyeballing chromatograms in the lab because, unlike the game, there's no feedback. If some machine could give the correct value, they wouldn't be eyeballing the chromatograms in the first place.

Oh, my best after 4 or 5 tries was about 3.5. I improved each round but I think it was mostly because I was concentrating harder and taking more time.

DBrower said...

Cub,

When doing manual integration, there are at least three kinds of feedback being given.

(1) visual indication on the peak;
(2) visual indication on the 2/1 trace also displayed;
(3) a current delta value displayed as you move the integration points.

There was testimony on both sides of whether the 2/1 trace is a reliable indicator for these purposes. One view is that it helps, as long as the peaks are uncontaminated -- this boils down to how good the separation chemistry was, and how appropriate the chromatographic separation was.

More troublesome is that, as we showed in the integration for idiots series, minor changes in limits can have large effects on the reported delta. If one goes into manual integration with preconceptions or inclinations about what the value should be, and one is seeing values displayed, one might easily skew the integration to achieve the expected values. We need not assign nefarious motive to this; it can be completely subconscious.

Imagine playing the eyeball game here with a "result" field displayed while you are placing the point, and it has a random error of 1.5 units in one of the four directions. Are you going to use your eyeballs on the figure, or try to get the error value zeroed? If you ignore the number, you'll probably be more accurate; if you don't, your score is likely to be 1.5 worse than you'd have gotten without the incorrect feedback.

TBV