2
$\begingroup$

I am slightly confused on a paper I am reading. They have a table of calculated Pearson coefficients as shown here:

$$ \begin{array} . & I & II & III\\ A + B & 0.8 & 0.65 & 0.90\\ A + C & 0.73 & 0.46 & 0.83\\ B + C & 0.61 & 0.76 & 0.68\\ \end{array} $$

They claim that the weakest correlation recorded was between $B+C, \,\, I$ and $B+C, \,\, III$. $A,B,C$ are different imaging modalities.

Are they stating that these are the two closest values in one comparison (i.e $B+C$)?

I am confused because they are now correlating correlations!

I thought the the weakest correlations would be $B+C, \,\, I$ and $A+C, \,\, II$, as they have the smallest values, but perhaps I did not understand their statement.

The paper can be found here: http://online.medphys.org/resource/1/mphya6/v40/i4/p041707_s1?bypassSSO=1

Page 4, section III.B

$\endgroup$
  • $\begingroup$ Can you give a reference, or at least some kind of direct quote so we can see what they actually say? $\endgroup$ – Glen_b Mar 19 '13 at 1:40
  • $\begingroup$ I have added the URL to the paper. Thank you for your time. $\endgroup$ – Shinobii Mar 19 '13 at 1:52
  • $\begingroup$ I'm looking at the html version, so page numbers don't help. If you mean the figures in Table II, the correlations there don't match what you have typed in. $\endgroup$ – Glen_b Mar 19 '13 at 2:01
  • $\begingroup$ Yes, that is the table. They have made errors (one I believe 0.5 should be 0.65). I do not know how this paper got published, lots of errors. $\endgroup$ – Shinobii Mar 19 '13 at 2:07
  • 1
    $\begingroup$ The presence of errors in the values makes it hard to say much of anything because their comments may be based on numbers other than the ones either of us are looking at. Maybe they made yet another table between the original figures and the published table and their comments refer to that - if they are so careless, how could we tell what else may have happened? You'd need to ask the authors. $\endgroup$ – Glen_b Mar 19 '13 at 3:01
2
$\begingroup$

The presence of errors in the values makes it hard to say much of anything because their comments may be based on numbers other than the ones either of us are looking at. Maybe they made yet another table between the original figures and the published table and their comments refer to that - if they are so careless, how could we tell what else may have happened? You'd need to ask the authors.

would their statement be correct? Is it correct to say that if two correlations are close in value (within a given group) that they have a weak correlation? As in the table above, the smallest difference between two correlations is 0.61-0.69 as they state?

The comment about weak correlation would presumably be because the correlation values themselves aren't large; you can't pick two correlations that are similar in size and thereby claim that the correlations are themselves correlated.

There may be an argument that some of the correlations are in turn correlated, but an argument based on their values doesn't establish that.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.