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 (January 2008)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Tue, 29 Jan 2008 16:30:30 +0100
Reply-To:     Marta García-Granero <mgarciagranero@gmail.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Marta García-Granero <mgarciagranero@gmail.com>
Subject:      Re: Conditional logistic problems
In-Reply-To:  <002701c861ea$4adf5150$2845cd80@ssw.buffalo.edu>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Hi Gene:

See below a fully worked example, solved using the three approaches (McNemar's OR with a MACRO, Mantel-Haenszel estimator and Conditional logistic regression).

HTH, Marta

(don't hesitate to ask me any question you might have concerning the syntax)

* Sample dataset (1:1 matched binary data, lower value means exposed in both groups) *. DATA LIST FREE/Case Control (2 F8). BEGIN DATA 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 END DATA. VALUE LABEL Case Control 1'Exposed to Risk' 2 'Non exposed'. LIST /CASES=FROM 1 TO 10.

* 1rst approach, valid only for 1:1 matched data, using a MACRO *. * MACRO call (see my message to the list: "Re: Help with Matrix output (& Paired data OR macro too, for H)"*.

PAIREDOR Case Control.

* 2nd approach: Mantel-Haenszel stratified analysis (valid for 1:1 or 1:k matched data) *.

* First of all, data need restructuring (using VARSTOCASES) *. VARSTOCASES /ID = pair /MAKE Exposition 'Exposed to risk factor' FROM Case Control /INDEX = Outcome(2). VALUE LABEL Outcome 1'Case' 2 'Control'.

* MH OR (ignore all tables but the last) *. CROSSTABS /TABLES=Exposition BY Outcome BY pair /FORMAT=NOTABLES /STATISTIC=CMH(1) /COUNT ROUND CELL .

* 3rd approach: Conditional logistic regression (thru Cox regression) *. * We will start from the dataset we have just restructured *. * We need three variables: - Exposition -> it is already OK - Pseudotime: should have these two values: 1 for cases and 2 for controls we already have it (named outcome). - Status: should have these two values: 0 for controls and 1 for cases we must create it.

COMPUTE status=(outcome=1).

COXREG Outcome /STATUS=status(1) /STRATA=pair /CONTRAST (Exposition)=Indicator /METHOD=ENTER Exposition /PRINT=CI(95).

The three methods give the same result (as expected). > All, > > I have been working with another list subscriber who has a matched (1:1) > dataset and is trying to analyze relationships between variables using > conditional logistic regression (CLR). I have no experience with this type > of model. However, I did find a posting by Marta (shown below) as well as > several discussion sites. Since spss logistic won't do a conditinal > analysis, the trick is to use the coxreg procedure to do so. I believe I > have faithfully followed Marta's directions but am getting no results and > so, there is something I don't understand. Per Marta's posting, I have > > COXREG ftime /STATUS=outcome(1) /STRATA=pair > /METHOD=ENTER x /PRINT=CI(95). > > Where > Ftime is the survival time variable computed such that ftime=outcome+1. > outcome is the case-control variable with 0=control, 1=case. > Pair is the strata variable, 161 pairs. > X is the IV. > > My case processing box shows > > Case Processing Summary > N Percent > Cases available in analysis Event(a) 161 50.0% > Censored 0 .0% > Total 161 50.0% > Cases dropped Cases with missing values 0 .0% > Cases with negative time 0 .0% > Censored cases before the > earliest event in a stratum 161 50.0% > Total 322 100.0% > a Dependent Variable: ftime > > Variables in the Equation(b) > Wald df Sig. > X . 0(a) . > aDegree of freedom reduced because of constant or linearly dependent > covariates > bConstant or Linearly Dependent Covariates > S = Stratum effect. x = .5093 + S; > > > If somebody can educate me about what I'm missing or need to look at, I'd > greatly appreciate it. > > Thanks, Gene Maguin > > ===================== > To manage your subscription to SPSSX-L, send a message to > LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the > command. To leave the list, send the command > SIGNOFF SPSSX-L > For a list of commands to manage subscriptions, send the command > INFO REFCARD > >

===================== To manage your subscription to SPSSX-L, send a message to LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD


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