LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (March 2002)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:   Tue, 19 Mar 2002 15:26:18 -0300
Sender:   "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:   Hector Maletta <>
Subject:   Re: "r" cutoff
Content-Type:   text/plain; charset=us-ascii

Dear Diane,

your physician boss is right, in my view, especially if you're concerned with health research as in your depression correlations. Suppose you get r=-0.30 between depression and job satisfaction, with p<.0001. This means that -0.30 squared=0.09 or 9% of the variation in depression scores is explained by job satisfaction (or insatisfaction, I presume it's preferable to say). Assuming the confounding influence of other factors has been controlled away, this is not a minor discovery. For instance, it tells you that 91% of differences in depression scores are unrelated to job satisfaction; i.e. giving everyone a satisfying job would reduce the variability of depression scores by only 9%. In terms of regression coefficients beta (deducible from r and the variables' variances), it tells you what reduction in depression scores you can expect if you improve job satisfaction scores in a given measure. This may be translated into millions of dollars saved by employers and HMO due to lower depression cases among employees, and this benefit compared to the cost of improving job satisfaction determining factors (whose effects on job satisfaction are determined by a different regression equation). Of course, life would be easier if a single factor would explain 100% of your problem, but usually many factors intervene, and thus amny of them would explain only a fraction of the total. If the estimate of that fraction has a low margin of error, the conclusions thereof may be useful in spite of the small percentage of variance explained.

Hector Maletta Universidad del Salvador Buenos Aires, Argentina

"Cohen,Diane" wrote: > > Dear Listers, > > I have run some social science correlations, all of which are significant (p=<.001). They have to do with correlations between total depression scores and job satisfaction scores, anxiety scores and life satisfaction scores, etc. I now need to know what the 'r' cut off is in determining which of these correlations are really worth reporting. My business major son says r's below .45 are not very interesting or boss (a physician) says all are worthwhile - whether .131 or .777. What do you social science folks say?? > > Diane

Back to: Top of message | Previous page | Main SPSSX-L page