In talk.politics.drugs Rich Ulrich <[EMAIL PROTECTED]> wrote: > On 14 Dec 2002 05:30:59 -0600, Brian Sandle > <[EMAIL PROTECTED]> wrote:
> [snip, various stuff] >> >> But if you were hurrying, and just looked at the calculation example for >> Pearson product moment correlation, you would not see the *assumptions* >> you make when using it. Yes you would see that you use it when you have >> actual scores attained by each subject, rather than just who is better >> than who. And you would see you were working with who is better than who, >> without knowing the scores, in Spearman's rank correlation. > To this statistician, the use here of the word 'assumptions' > seems unique. Yes, when we have certain data -- ranks -- we > get the Spearman when we compute the Pearson. > We can compute a Pearson correlation regardless of > assumptions about outliers, etc., Yes, just put in the ranks instead of scores. Of ocurse the machine computes away, it knows no better. But how does that have meaning? Are you saying you get exactly the same correlations when you put ranks in Pearson as you do when you put the same ones in Spearman? Isn't it better to be `regardful' of non-linear stuff and use the proper formula? and what we sacrifice > might be (a) the test where we can rely on p-levels, and In SPSS, if you were to do rank correlations, and use the Spearman, you would still have p-levels, i.e. significance. And I think you would still have p-levels if you did Spearman partials, that is with linear stuff &c. But if non-linearity forces you to go to partial rank, then you used to lose the p-levels. But iof you go stiking those ranks in Pearson it still trots out p-levels, not knowing any better. Do you see your mistake? > (b) a test on 'something else' - such as, ranks. How do you sacrfice a test on ranks? > Computers make this all so much simpler. But bring in traps. That is, I think I > now prefer to drop all the special names when the r's > being the usual Pearson. That is not a sentence. Instead, I talk about doing a > 'rank-transformation' to the data; and whether that is > desirable or useful or damaging. By that you are turning scores into ranks? You *have* to do that when the data is non-linear is my guess. Then you put it into a formula designed to work with ranks. Otherwise what you get out has no meaning. > [ snip] >> You use the Spearman test if you do not have scores, but if you have the >> scores, why bother with the Pearson test, because you can use the Spearman >> anyway, and if you are working by hand, it is about 1/3 the work. Why? >> More to it? > Beg-pardon -- Who works by hand? Right, and it was put in to indicate the great difference in the formulae for Pearson's and Spearman's. > I will second the suggestion that you drop the subject, or > at least, take sci.stat.edu off the line for newsgroups, if > there is not a statistical question. I don't understand you. Perhaps you should give some refs. -----= Posted via Newsfeeds.Com, Uncensored Usenet News =----- http://www.newsfeeds.com - The #1 Newsgroup Service in the World! -----== Over 80,000 Newsgroups - 16 Different Servers! =----- . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
