# Results from clogit out of range?

 Classic List Threaded
8 messages
Reply | Threaded
Open this post in threaded view
|

## Results from clogit out of range?

 Hi, I'm calculating a conditional logit on some data stratified by group. My understanding was that a conditional logit by definition returns a value between 0 and 1 a a probability.  Can anyone suggest why I'm seeing results outside of the {0,1} range?? The call in R is: m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) Then prediction <- predict(m,newdata) A sample of the data with resulting predictive values is:      group          score   val_1               val_2        prediction 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: > Hi, > > I'm calculating a conditional logit on some data stratified by group. > My understanding was that a conditional logit by definition returns a > value between 0 and 1 a a probability.  Can anyone suggest why I'm > seeing results outside of the {0,1} range?? Probably because you did not read the help page for the function   predict.coxph. Pay special attention to the type argument. "type=c("lp", "risk", "expected", "terms")" > > The call in R is: > > m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) > > Then > > prediction <- predict(m,newdata) I do not see that you have defined any newdata. > > > A sample of the data with resulting predictive values is: >     group          score   val_1               val_2        prediction > 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 > 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 > 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 > 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 > 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 > 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 > 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 > 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 > 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 > 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 > 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 By default (which is what you have implicitly chosen) you are   requesting "lp" = linear predictors rather than "risk". -- David. > > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. David Winsemius, MD West Hartford, CT ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 Thanks David, That explains a lot.  I appreciate it. -- Noah On 4/20/10 3:48 PM, David Winsemius wrote: > > On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: > >> Hi, >> >> I'm calculating a conditional logit on some data stratified by group. >> My understanding was that a conditional logit by definition returns a >> value between 0 and 1 a a probability.  Can anyone suggest why I'm >> seeing results outside of the {0,1} range?? > > Probably because you did not read the help page for the function > predict.coxph. Pay special attention to the type argument. > > "type=c("lp", "risk", "expected", "terms")" > >> >> The call in R is: >> >> m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) >> >> Then >> >> prediction <- predict(m,newdata) > > I do not see that you have defined any newdata. > >> >> >> A sample of the data with resulting predictive values is: >>     group          score   val_1               val_2        prediction >> 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 >> 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 >> 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 >> 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 >> 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 >> 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 >> 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 >> 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 >> 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 >> 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 >> 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 > > By default (which is what you have implicitly chosen) you are > requesting "lp" = linear predictors rather than "risk". > ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 I just read the help page for predict.coxph. It indicates that the risk score is just exp(lp) What I'm trying to find, and have seen with some other implementations is the "conditional probability within group".  Neither the lp or the risk options seem to deliver this. What am I missing? -N On 4/20/10 4:22 PM, Noah Silverman wrote: > Thanks David, > > That explains a lot.  I appreciate it. > > -- > Noah > > > On 4/20/10 3:48 PM, David Winsemius wrote: >   >> On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: >> >>     >>> Hi, >>> >>> I'm calculating a conditional logit on some data stratified by group. >>> My understanding was that a conditional logit by definition returns a >>> value between 0 and 1 a a probability.  Can anyone suggest why I'm >>> seeing results outside of the {0,1} range?? >>>       >> Probably because you did not read the help page for the function >> predict.coxph. Pay special attention to the type argument. >> >> "type=c("lp", "risk", "expected", "terms")" >> >>     >>> The call in R is: >>> >>> m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) >>> >>> Then >>> >>> prediction <- predict(m,newdata) >>>       >> I do not see that you have defined any newdata. >> >>     >>> >>> A sample of the data with resulting predictive values is: >>>     group          score   val_1               val_2        prediction >>> 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 >>> 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 >>> 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 >>> 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 >>> 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 >>> 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 >>> 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 >>> 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 >>> 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 >>> 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 >>> 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 >>>       >> By default (which is what you have implicitly chosen) you are >> requesting "lp" = linear predictors rather than "risk". >> >>     > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. > ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 On Apr 20, 2010, at 7:31 PM, Noah Silverman wrote: > I just read the help page for predict.coxph. > > It indicates that the risk score is just exp(lp) > > What I'm trying to find, and have seen with some other implementations > is the "conditional probability within group".  Neither the lp or the > risk options seem to deliver this. > > What am I missing? > I'm not sure that clogit lets you estimate that quantity: http://tolstoy.newcastle.edu.au/R/help/06/02/21443.htmlDo you have citations that back up your original assumption? --   David, > -N > > > On 4/20/10 4:22 PM, Noah Silverman wrote: >> Thanks David, >> >> That explains a lot.  I appreciate it. >> >> -- >> Noah >> >> >> On 4/20/10 3:48 PM, David Winsemius wrote: >> >>> On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: >>> >>> >>>> Hi, >>>> >>>> I'm calculating a conditional logit on some data stratified by   >>>> group. >>>> My understanding was that a conditional logit by definition   >>>> returns a >>>> value between 0 and 1 a a probability.  Can anyone suggest why I'm >>>> seeing results outside of the {0,1} range?? >>>> >>> Probably because you did not read the help page for the function >>> predict.coxph. Pay special attention to the type argument. >>> >>> "type=c("lp", "risk", "expected", "terms")" >>> >>> >>>> The call in R is: >>>> >>>> m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) >>>> >>>> Then >>>> >>>> prediction <- predict(m,newdata) >>>> >>> I do not see that you have defined any newdata. >>> >>> >>>> >>>> A sample of the data with resulting predictive values is: >>>>    group          score   val_1               val_2         >>>> prediction >>>> 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 >>>> 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 >>>> 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 >>>> 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 >>>> 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 >>>> 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 >>>> 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 >>>> 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 >>>> 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 >>>> 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 >>>> 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 >>>> >>> By default (which is what you have implicitly chosen) you are >>> requesting "lp" = linear predictors rather than "risk". >>> >>> >> ______________________________________________ >> [hidden email] mailing list >> https://stat.ethz.ch/mailman/listinfo/r-help>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html>> and provide commented, minimal, self-contained, reproducible code. >> > > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. David Winsemius, MD West Hartford, CT ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 In reply to this post by Noah Silverman-3 On Tue, 20 Apr 2010, Noah Silverman wrote: > I just read the help page for predict.coxph. > > It indicates that the risk score is just exp(lp) > > What I'm trying to find, and have seen with some other implementations > is the "conditional probability within group".  Neither the lp or the > risk options seem to deliver this. exp(lp)/(1+exp(lp)) or, as Martin Maechler would prefer,  qlogis(lp)     -thomas > What am I missing? > > -N > > > On 4/20/10 4:22 PM, Noah Silverman wrote: >> Thanks David, >> >> That explains a lot.  I appreciate it. >> >> -- >> Noah >> >> >> On 4/20/10 3:48 PM, David Winsemius wrote: >> >>> On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: >>> >>> >>>> Hi, >>>> >>>> I'm calculating a conditional logit on some data stratified by group. >>>> My understanding was that a conditional logit by definition returns a >>>> value between 0 and 1 a a probability.  Can anyone suggest why I'm >>>> seeing results outside of the {0,1} range?? >>>> >>> Probably because you did not read the help page for the function >>> predict.coxph. Pay special attention to the type argument. >>> >>> "type=c("lp", "risk", "expected", "terms")" >>> >>> >>>> The call in R is: >>>> >>>> m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) >>>> >>>> Then >>>> >>>> prediction <- predict(m,newdata) >>>> >>> I do not see that you have defined any newdata. >>> >>> >>>> >>>> A sample of the data with resulting predictive values is: >>>>     group          score   val_1               val_2        prediction >>>> 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 >>>> 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 >>>> 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 >>>> 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 >>>> 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 >>>> 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 >>>> 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 >>>> 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 >>>> 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 >>>> 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 >>>> 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 >>>> >>> By default (which is what you have implicitly chosen) you are >>> requesting "lp" = linear predictors rather than "risk". >>> >>> >> ______________________________________________ >> [hidden email] mailing list >> https://stat.ethz.ch/mailman/listinfo/r-help>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html>> and provide commented, minimal, self-contained, reproducible code. >> > > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. > Thomas Lumley Assoc. Professor, Biostatistics [hidden email] University of Washington, Seattle ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 Thomas Lumley wrote: > On Tue, 20 Apr 2010, Noah Silverman wrote: > >> I just read the help page for predict.coxph. >> >> It indicates that the risk score is just exp(lp) >> >> What I'm trying to find, and have seen with some other implementations >> is the "conditional probability within group".  Neither the lp or the >> risk options seem to deliver this. > > exp(lp)/(1+exp(lp)) > > or, as Martin Maechler would prefer,  qlogis(lp) Minor correction: plogis Frank > >    -thomas > > >> What am I missing? >> >> -N >> >> >> On 4/20/10 4:22 PM, Noah Silverman wrote: >>> Thanks David, >>> >>> That explains a lot.  I appreciate it. >>> >>> -- >>> Noah >>> >>> >>> On 4/20/10 3:48 PM, David Winsemius wrote: >>> >>>> On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: >>>> >>>> >>>>> Hi, >>>>> >>>>> I'm calculating a conditional logit on some data stratified by group. >>>>> My understanding was that a conditional logit by definition returns a >>>>> value between 0 and 1 a a probability.  Can anyone suggest why I'm >>>>> seeing results outside of the {0,1} range?? >>>>> >>>> Probably because you did not read the help page for the function >>>> predict.coxph. Pay special attention to the type argument. >>>> >>>> "type=c("lp", "risk", "expected", "terms")" >>>> >>>> >>>>> The call in R is: >>>>> >>>>> m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) >>>>> >>>>> Then >>>>> >>>>> prediction <- predict(m,newdata) >>>>> >>>> I do not see that you have defined any newdata. >>>> >>>> >>>>> >>>>> A sample of the data with resulting predictive values is: >>>>>     group          score   val_1               val_2        prediction >>>>> 1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 >>>>> 2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 >>>>> 3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 >>>>> 4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 >>>>> 5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 >>>>> 6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 >>>>> 7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 >>>>> 8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 >>>>> 9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 >>>>> 10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 >>>>> 11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 >>>>> >>>> By default (which is what you have implicitly chosen) you are >>>> requesting "lp" = linear predictors rather than "risk". > Thomas Lumley            Assoc. Professor, Biostatistics > [hidden email]    University of Washington, Seattle -- Frank E Harrell Jr   Professor and Chairman        School of Medicine                       Department of Biostatistics   Vanderbilt University ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code. Frank Harrell Department of Biostatistics, Vanderbilt University
Reply | Threaded
Open this post in threaded view
|

## Re: Results from clogit out of range?

 On Wed, 21 Apr 2010, Frank E Harrell Jr wrote: > Thomas Lumley wrote: >>  On Tue, 20 Apr 2010, Noah Silverman wrote: >> >> >  I just read the help page for predict.coxph. >> > >> >  It indicates that the risk score is just exp(lp) >> > >> >  What I'm trying to find, and have seen with some other implementations >> >  is the "conditional probability within group".  Neither the lp or the >> >  risk options seem to deliver this. >> >>  exp(lp)/(1+exp(lp)) >> >>  or, as Martin Maechler would prefer,  qlogis(lp) > > Minor correction: plogis In what context is this reasonble? It seems to me that rather strong assumptions are needed - like that the strata are independent of the covariate(s). The OP requested the "conditional probability within group", which I interprete to mean something like   exp(lp)/ave( exp(lp), stratvar, FUN=sum ) when there is just one case in each stratum. ?? Chuck > > Frank > >> >>     -thomas >> >> >> >  What am I missing? >> > >> >  -N >> > >> > >> >  On 4/20/10 4:22 PM, Noah Silverman wrote: >> > >  Thanks David, >> > > >> > >  That explains a lot.  I appreciate it. >> > > >> > >  -- >> > >  Noah >> > > >> > > >> > >  On 4/20/10 3:48 PM, David Winsemius wrote: >> > > >> > > >  On Apr 20, 2010, at 5:59 PM, Noah Silverman wrote: >> > > > >> > > > >> > > > >  Hi, >> > > > > >> > > > >  I'm calculating a conditional logit on some data stratified by >> > > > >  group. >> > > > >  My understanding was that a conditional logit by definition >> > > > >  returns a >> > > > >  value between 0 and 1 a a probability.  Can anyone suggest why I'm >> > > > >  seeing results outside of the {0,1} range?? >> > > > > >> > > >  Probably because you did not read the help page for the function >> > > >  predict.coxph. Pay special attention to the type argument. >> > > > >> > > >  "type=c("lp", "risk", "expected", "terms")" >> > > > >> > > > >> > > > >  The call in R is: >> > > > > >> > > > >  m <- clogit(score ~ val_1 + val_2 + strata(group), data=data) >> > > > > >> > > > >  Then >> > > > > >> > > > >  prediction <- predict(m,newdata) >> > > > > >> > > >  I do not see that you have defined any newdata. >> > > > >> > > > >> > > > > >> > > > >  A sample of the data with resulting predictive values is: >> > > > >      group          score   val_1               val_2 >> > > > >  prediction >> > > > >  1  2009-01-04_1     1 0.5913962     -1.121589  1.62455210 >> > > > >  2  2009-01-04_1     1 0.6175472     -3.249820 -0.20093346 >> > > > >  3  2009-01-04_1     1 0.5439640     -2.424501  0.46651849 >> > > > >  4  2009-01-04_1     0 0.3745209     -2.477424  0.31263855 >> > > > >  5  2009-01-04_1     0 0.6329855     -3.424174 -0.34200448 >> > > > >  6  2009-01-04_1     0 0.4571999     -2.770247  0.11190788 >> > > > >  7  2009-01-04_1     0 0.3822623     -2.259422  0.50627534 >> > > > >  8  2009-01-04_1     0 0.2605742     -4.424806 -1.44566070 >> > > > >  9  2009-01-04_1     0 0.4449604     -2.357060  0.46174993 >> > > > >  10 2009-01-04_1     0 0.6595178     -2.246518  0.69427842 >> > > > >  11 2009-01-04_1     0 0.5260032     -2.977887 -0.02393871 >> > > > > >> > > >  By default (which is what you have implicitly chosen) you are >> > > >  requesting "lp" = linear predictors rather than "risk". > >>  Thomas Lumley            Assoc. Professor, Biostatistics >>  [hidden email]    University of Washington, Seattle > > > -- > Frank E Harrell Jr   Professor and Chairman        School of Medicine >                      Department of Biostatistics   Vanderbilt University > > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. > Charles C. Berry                            (858) 534-2098                                              Dept of Family/Preventive Medicine E mailto:[hidden email]            UC San Diego http://famprevmed.ucsd.edu/faculty/cberry/  La Jolla, San Diego 92093-0901 ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.