Re: How to show that r-squared is not the be all and end all?

<Felicis@xxxxxxxxx> wrote in message
news:1155139408.667109.173800@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
What is the difference between the two R^2 values? If one is (for
example) 0.7 and the other 0.99, then I would say that the higher one
is definitely better, but if one is 0.7235 and the other 0.7236, then
there is essentially no difference in correlation.

The fund has high correlations to both benchmarks. This is not surprising
since all of the stocks the fund owns is included in both benchmarks and one
benchmark makes up 75% of the other benchmark (i.e. the Russell Top 200
represents 75% of the Russell 1000 with the remaining 25% coming from the
Russell Midcap).

If both R^2 values are fairly low, then the fit is poor in both cases,
and you might be better served in trying to find some other model.

Remember also that past performance is no guarantee of future success -
just because there is a correlation to date does not mean there will be
one in the future.

I took a moment to look up some things on wikipedia:

There is an interesting note towards the bottom of this one showing
four data sets that all have the same R^2, (and the same best-fit
linear regression line), but completely different curves, showing the
limits of R^2 as a measurement tool (which is what you seem to be
interested in).

Thanks for the links. I did not see the information at the bottom of the R
Squared link with the four data sets that have different curves. I checked
several of the links from that page and did not see it there either. Are
you sure it was on the same page as the link you posted?


.

Relevant Pages

  • Re: Another EPR question
    ... The conclusion that the coincidence curves ... detail to account for the correlation curves, ... >> understanding of polarization than any sort of instantaneous ...
    (sci.physics.research)
  • Re: entanglement question
    ... Measurement of 's property would confirm ... The correlation is enforced even across ... Roll a die. ... The top-number is just as random as the bottom ...
    (sci.physics)
  • Re: What is a (sci.crypt expert) ?
    ... contrivance to get Smug Doug's name in the Beale cipher faq? ... Basically it works like becoming an expert elsewhere. ... Keep showing up, keep being the best, keep getting better, keep pushing further, and eventually you are recognised as a respected expert. ... There also seems to be a roughly inverse correlation between frequency and respect, for example Coppersmith posted once, Austin Osbyrne won't shut up, there are exceptions, but more respected people tend to only speak when they have something to genuinely say and contribute. ...
    (sci.crypt)
  • Re: Poll OT Toilet paper
    ... They say you use less if you pull it over the bottom than over the top is ... I haven't found a correlation between kitty intelligence and toilet ...
    (rec.pets.cats.anecdotes)