PDA

View Full Version : Statistical analysis of Track world records


Southern fried XC
01-09-2011, 10:49 PM
For any of you all with some background in statistics, I was doing a small project for my AP stat class and I came across some interesting information

For almost all of the track world record progressions (world record time plotted against time), theres a really strong association. Theres a level of ambiguity to the line of best fit, but the correlation coefficient for the 10k for example is -.9876.

I think this is startling and almost counter intuitive; one would think there are apparent limits in the human body and this would suggest perhaps an exponential model would fit well but a residual plot shows nothing too striking regarding the validity of the regression.

Another thing, a residual plot shows that the 1950s-1980s were (in relative terms) the most successful years of track and field; records set in those years deviated from the line of best fit by around -10 seconds.

If you want the data (its a bit of a hassle to look at each progression), let me know

808runner
01-12-2011, 03:21 AM
I think this is startling and almost counter intuitive; one would think there are apparent limits in the human body and this would suggest perhaps an exponential model would fit well but a residual plot shows nothing too striking regarding the validity of the regression.

Another thing, a residual plot shows that the 1950s-1980s were (in relative terms) the most successful years of track and field; records set in those years deviated from the line of best fit by around -10 seconds.



Can you send me graphs of your residuals? To me, if there is a chunk of data above/below the residual, then chances are your data doesn't fit.
Remember, residuals should show no pattern whatsoever. No smiles, frowns, spreading. Just random plotting.

Show'EmTheRopes
01-12-2011, 11:46 AM
I did something similar with the progressions using Calculus for my AP Project (using every WR I could fine for each event). Most of the progressions had a line of best fit similar to a portion of a Gaussian "Bell" curve. From this, it was easy to make a reasonable prediction about the fastest a WR could possibly get.

I wish I still had all my data and project stuff... I remember the 100m getting down to 9.39 and 200m down to 18.6 or something like that.

backtoback
01-25-2011, 09:30 PM
http://www.boston.com/bostonglobe/ideas/articles/2011/01/23/peaked_performance/?page=full

Interesting read and somewhat related