home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!sdd.hp.com!ncr-sd!npg-sd!rns
- From: rns@npg-sd.SanDiegoCA.NCR.COM (Rick Schubert)
- Newsgroups: rec.running
- Subject: Re: 10k-half-marathon
- Message-ID: <5667@npg-sd.SanDiegoCA.NCR.COM>
- Date: 20 Aug 92 19:43:15 GMT
- References: <46412@shamash.cdc.com> <1992Aug14.092745.19779@nestroy.wu-wien.ac.at> <1992Aug14.133401.28594@bmers145.bnr.ca> <1992Aug14.164038.10053@gdr.bath.ac.uk> <92231.083017RBPQCC@rohvm1.rohmhaas.com>
- Organization: NCR Corp., Network Products - San Diego
- Lines: 34
-
- In <92231.083017RBPQCC@rohvm1.rohmhaas.com> RBPQCC@rohvm1.rohmhaas.com writes:
-
- >Still another way...
- >Plot your race times vs race distance on log-log paper. If all were
- >equal effort they will form a (more or less) straight line with a slope
- >slightly greater than 1. Now you can extrapolate/interpolate to any
- >distance you want. It definitely breaks down past the marathon.I am
- >fairly linear from 1 mile up to marathon. It is also good to point
- >out your "soft" PRs.
-
- I think the main reason that it "breaks down past the marathon" is that
- you are usually extrapolating there rather than interpolating. Extrapolation
- is usually less reliable than interpolation. I suspect that if you
- have data points for 10K, marathon, and 100 miles, that the interpolation
- of your 50 mile time would be pretty reasonable.
-
- The same problem should occur to extrapolating to shorter distances.
- I suspect that if you had data points of 5K, 10K, half marathon, and
- marathon, that extrapolating to 100 meters would "break down" also.
-
- The other reason that extrapolation/interpolation may break down is that
- a linear approximation assumes that you are equally-good at all distances.
- So, even if you had a data point for 100 miles, but were only trained
- for the marathon, interpolating a 50-mile time might not be too accurate
- if you walked the last 50 miles of the 100-mile race.
-
- Has anyone seen any models that add an extra parameter that can account
- for ones "best" distance? It would be interesting to try to develop
- such a model given enough data.
-
- -- Rick Schubert (rick.schubert@SanDiegoCA.NCR.COM)
-
- Disclaimer: I haven't had a statistics course in about 17 years, so I may
- be a bit rusty.
-