Re: Using correlation coefficient
From: Predictor (predictr_at_bellatlantic.net)
Date: 12/22/04
- Next message: Osher Doctorow: "Re: Phases Underlie Probability-Statistics: PI, FFE, VFE."
- Previous message: nantipov: "Re: A NEW Electronic BOOK on STATISTICS!"
- Maybe in reply to: generic_at_arigatoo.net: "Using correlation coefficient"
- Messages sorted by: [ date ] [ thread ]
Date: 22 Dec 2004 11:59:46 -0800
beliavsky@aol.com wrote:
> Bruce Weaver <bweaver@lakeheadu.ca> wrote:
> >generic@arigatoo.net wrote:
> >
> >--- snip -----
> >
> >> Here are the results of this example using Excel.
> >>
> >>
> >> Pred 1 Correlation Coeffic: .8
> >> Pred 2 Correlation Coeffic: 1
> >>
> >
> >Here is output from SPSS for the same data you used:
> >
> >ID P1 P2 ACTUAL
> >
> > 1 1 1 1
> > 2 3 2 2
> > 3 2 3 3
> > 4 4 4 4
> >
> >Number of cases read: 4 Number of cases listed: 4
> >
> >CORRELATIONS
> > /VARIABLES=p1 p2 actual
> > /PRINT=TWOTAIL NOSIG
> > /MISSING=PAIRWISE .
> >
> >Correlations
> >
> >|----------------------|----|---------|---------|
> >| |P1 |P2 |ACTUAL |
> >|------|---------------|----|---------|---------|
> >|P1 |Pearson |1 |.800 |.800 |
> >| |Correlation | | | |
> >|------|---------------|----|---------|---------|
> >|P2 |Pearson |.800|1 |1.000(**)|
> >| |Correlation | | | |
> >|------|---------------|----|---------|---------|
> >|ACTUAL|Pearson |.800|1.000(**)|1 |
> >| |Correlation | | | |
> >|------|---------------|----|---------|---------|
> >¢
> >** Correlation is significant at the 0.01 level (2-tailed).
> >
> >
> >So, to answer one of your questions, it appears you are using the
> >correlation function in Excel properly. (But note that the accuracy
of
>
> >statistical functions in Excel has been called into question. You'd
be
>
> >better off using a bona fide stats package if one is available.)
> >
> >
> >> My questions:
> >>
> >>
> >> - Is it accurate to say the following?
> >> "Prediction 1 had a strong correlation to the actual rankings."
> >> "Prediction 2 had a very strong correlation to the actual
rankings."
> >
> >In the data set you used, Prediction 2 has a perfect (positive)
linear
> >relationship with the actual ranking, which is stronger than "very
> >strong" in my book.
> >
> >>
> >>
> >> - Am I using this function correctly?
> >> - Anything you would recommend to firm up this test?
> >
> >
> >Are you using ranks because that's all you have? Or are you
converting
>
> >actual scores to ranks? If the latter, you could well be better off
> >using the actual data. Ranking throws away information.
>
> True, but the Pearson correlation measure you are talking about is
not robust
> to bad data. The Spearman rank correlation that the OP is computing
is more
> robust. I think both measures are worth computing.
Have you calculated the Spearman rank-order correlation? What did you
find?
-Will Dwinnell
http://will.dwinnell.com
- Next message: Osher Doctorow: "Re: Phases Underlie Probability-Statistics: PI, FFE, VFE."
- Previous message: nantipov: "Re: A NEW Electronic BOOK on STATISTICS!"
- Maybe in reply to: generic_at_arigatoo.net: "Using correlation coefficient"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
