Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



Correlation coefficient question
Posted:
Apr 3, 1995 1:47 PM


I'd like to pose a question to the group regarding the correlation coefficient, r. As many of you may know, the TI calculator products display r as part of the output with several of the regression models: linreg, lnreg, pwrreg, etc. Note that we do not do any r correlation for the quadratic, cubic, and quartic models?
I have heard pros and cons about the display of r  some prefer that it be displayed with each of these models and some prefer that it not be displayed, but that it be "hidden" away under a variables screen. By hiding it away, the opinion is, we are keeping people from misusing it that are misinterpreting its meaning.
My question is: What do you think? Should r be displayed with the models or should it be "hidden"? Would it be better to display r^2 rather than r? Or should we display both? Is there some other measure of fit that would be better?
Thanks for your input, Charlotte Clark cclark@ti.com



