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 (February 2008, week 1)Back to main SAS-L pageJoin or leave SAS-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Sun, 3 Feb 2008 22:21:59 -0800
Reply-To:     John Amora <bayesian2001@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         John Amora <bayesian2001@YAHOO.COM>
Subject:      Re: logistic regression
Comments: To: "Dennis G. Fisher, Ph.d." <dfisher@CSULB.EDU>
In-Reply-To:  <47A69E52.8000608@csulb.edu>
Content-Type: text/plain; charset=iso-8859-1

Consider the estimated logistic regression equation with p predictor variables:, log odds = B_0 + B_1X_1 + B_2*X_2+ ...+ B_p*X_p. That is, in logistic regression, the "log odds" is equal to the linear combination of the independent variables with B's as unstandardized coefficients. Since the left side is "log odds", the B's do not make sense. The intuitive ones are the exp(B); thus, most stat packages like SAS or SPSS provide exp(B) in their computer printouts. I dont think standardized B's are intuitive! Please check if the term BETA in the APA book refers really to standardized B.

Johnny

--- "Dennis G. Fisher, Ph.d." <dfisher@CSULB.EDU> wrote:

> Dan, > What I am confused about it that it seems as though > the only reasonable > thing to report is the > raw and not the standardized coefficients. However, > I have a book on > how to report statistics > in APA format and the chapter that they have for > logistic regression > shows them only reporting beta's which > I assume are the standardized coefficients. This > does not make sense. > The book should be showing > reporting B's instead. If you run the STB option, > it is possible to not > get coefficients at all for some of the > variables. This would make a lousy table for a > paper. I hope that > answers your question. I am not > sure that my answer is of any value, but there it > is. Thanks for your help. > Dennis > > Daniel Nordlund wrote: > >> -----Original Message----- > >> From: SAS(r) Discussion > [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Dennis > >> G. Fisher, Ph.d. > >> Sent: Sunday, February 03, 2008 10:50 AM > >> To: SAS-L@LISTSERV.UGA.EDU > >> Subject: logistic regression > >> > >> Are the "parameter estimates" that SAS prints in > PROC LOGISTIC > >> considered raw (reported as B's) or standardized > (reported as betas)? > >> Is there any option to convert whatever it prints > into the other? > >> Pampel (2000) says on page 32 of his book > "...logistic regression > >> programs do not routinely compute standardized > coefficients." However, > >> on page 33 he says "SAS prints standardized > coefficients for logistic > >> regression." However, he continues to criticize > the way that SAS > >> calculates these standardized coefficients > because SAS uses the same > >> standard deviation, 1.8138, for all variables. I > have not been able to > >> find this in the documentation yet. If someone > knows exactly where to > >> look in the documentation to clarify this, I > would appreciate their > >> pointing it out. I am trying to figure out how > to report these > >> parameter estimates correctly, whether as B's or > as beta's. TIA > >> Dennis Fisher > >> > >> -- > >> Dennis G. Fisher, Ph.D. > >> Professor and Director > >> Center for Behavioral Research and Services > >> California State University, Long Beach > >> 1090 Atlantic Avenue > >> Long Beach, California 90813 > >> 562-495-2330 x121 > >> 562-983-1421 fax > >> www.csulb.edu/centers/cbrs > >> > > > > As I understand the concept, SAS reports raw > coefficients. The reported coefficient for a > variable is the change in log(odds) based on a one > unit change (in the original scale) of the variable > of interest. Out of curiosity, what is your > interest in standardized coefficients? > > > > Hope this is helpful, > > > > Dan > > > > Daniel Nordlund > > Bothell, WA USA > > > > > > > > -- > Dennis G. Fisher, Ph.D. > Professor and Director > Center for Behavioral Research and Services > California State University, Long Beach > 1090 Atlantic Avenue > Long Beach, California 90813 > 562-495-2330 x121 > 562-983-1421 fax > www.csulb.edu/centers/cbrs >

____________________________________________________________________________________ Looking for last minute shopping deals? Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping


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