I am writing to get a better handle on a warning I am getting from a coxph analysis I am doing. I am analysing age of onset of dementia *after* the onset of parkinson disease. My data looks like: age.park age.dem age.death censor x1 x2 x3 x4 1 76 87 88 0 16 33 E3 E3 2 75 84 84 0 33 36 E3 E3 3 77 81 81 1 NA NA <NA> <NA> 4 65 65 69 0 NA NA E4 E4 5 56 76 79 0 NA NA <NA> <NA> 6 62 72 72 1 NA NA <NA> <NA> ... Obviously some individuals (lines 1,2,5) will first develop parkinson, then a few years later, dementia. Some individuals will not develop dementia (lines 3 and 6, where age of death and age of dementia correspond, but the censor variable is 1). Some (more) unluky individuals develop parkinson and dementia at the same time (line 4). my coxph model looks like coxph(Surv(age.mot,age.dem, censor) ~ x1 + x2 + x3 + x4, mydata) and I get the warning: In Surv(age.mot, age.dem, censor) : Stop time must be > start time, NA created I am almost sure that this is due to the instances where age.park == age.dem, but there is nothing I can really do. So my question: how do I deal with the instances where age.park == age.dem in order to keep those individuals in the analysis and to get sensible results? Best wishes Federico -- Federico C. F. Calboli Department of Epidemiology and Biostatistics Imperial College, St. Mary's Campus Norfolk Place, London W2 1PG Tel +44 (0)20 75941602 Fax +44 (0)20 75943193 f.calboli [.a.t] imperial.ac.uk f.calboli [.a.t] gmail.com
Surv(): Stop time must be > start time, NA created
4 messages · Federico Calboli, Dimitris Rizopoulos
this sounds like a competing risks problem. Maybe you would be interested in a cause-specific hazard regression or the Fine & Gray model (http://cran.r-project.org/package=cmprsk). Recently there was also a special issue in JSS on this topic (http://www.jstatsoft.org/v38). I hope it helps. Best, Dimitris
On 6/3/2011 12:17 PM, Federico Calboli wrote:
I am writing to get a better handle on a warning I am getting from a coxph analysis I am doing. I am analysing age of onset of dementia *after* the onset of parkinson disease. My data looks like: age.park age.dem age.death censor x1 x2 x3 x4 1 76 87 88 0 16 33 E3 E3 2 75 84 84 0 33 36 E3 E3 3 77 81 81 1 NA NA<NA> <NA> 4 65 65 69 0 NA NA E4 E4 5 56 76 79 0 NA NA<NA> <NA> 6 62 72 72 1 NA NA<NA> <NA> ... Obviously some individuals (lines 1,2,5) will first develop parkinson, then a few years later, dementia. Some individuals will not develop dementia (lines 3 and 6, where age of death and age of dementia correspond, but the censor variable is 1). Some (more) unluky individuals develop parkinson and dementia at the same time (line 4). my coxph model looks like coxph(Surv(age.mot,age.dem, censor) ~ x1 + x2 + x3 + x4, mydata) and I get the warning: In Surv(age.mot, age.dem, censor) : Stop time must be> start time, NA created I am almost sure that this is due to the instances where age.park == age.dem, but there is nothing I can really do. So my question: how do I deal with the instances where age.park == age.dem in order to keep those individuals in the analysis and to get sensible results? Best wishes Federico -- Federico C. F. Calboli Department of Epidemiology and Biostatistics Imperial College, St. Mary's Campus Norfolk Place, London W2 1PG Tel +44 (0)20 75941602 Fax +44 (0)20 75943193 f.calboli [.a.t] imperial.ac.uk f.calboli [.a.t] gmail.com
______________________________________________ R-help at r-project.org 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.
Dimitris Rizopoulos Assistant Professor Department of Biostatistics Erasmus University Medical Center Address: PO Box 2040, 3000 CA Rotterdam, the Netherlands Tel: +31/(0)10/7043478 Fax: +31/(0)10/7043014 Web: http://www.erasmusmc.nl/biostatistiek/
On 3 Jun 2011, at 11:27, Dimitris Rizopoulos wrote:
this sounds like a competing risks problem. Maybe you would be interested in a cause-specific hazard regression or the Fine & Gray model (http://cran.r-project.org/package=cmprsk).
I will look into that,but, biologically, parkinson leads to dementia (cause and effect rather than competing risk), and in fact all my subjects do have parkinson. Unless the two competing risks are 'risk of dementia' vs 'risk of death'. Best F
Recently there was also a special issue in JSS on this topic (http://www.jstatsoft.org/v38). I hope it helps. Best, Dimitris On 6/3/2011 12:17 PM, Federico Calboli wrote:
I am writing to get a better handle on a warning I am getting from a coxph analysis I am doing. I am analysing age of onset of dementia *after* the onset of parkinson disease. My data looks like: age.park age.dem age.death censor x1 x2 x3 x4 1 76 87 88 0 16 33 E3 E3 2 75 84 84 0 33 36 E3 E3 3 77 81 81 1 NA NA<NA> <NA> 4 65 65 69 0 NA NA E4 E4 5 56 76 79 0 NA NA<NA> <NA> 6 62 72 72 1 NA NA<NA> <NA> ... Obviously some individuals (lines 1,2,5) will first develop parkinson, then a few years later, dementia. Some individuals will not develop dementia (lines 3 and 6, where age of death and age of dementia correspond, but the censor variable is 1). Some (more) unluky individuals develop parkinson and dementia at the same time (line 4). my coxph model looks like coxph(Surv(age.mot,age.dem, censor) ~ x1 + x2 + x3 + x4, mydata) and I get the warning: In Surv(age.mot, age.dem, censor) : Stop time must be> start time, NA created I am almost sure that this is due to the instances where age.park == age.dem, but there is nothing I can really do. So my question: how do I deal with the instances where age.park == age.dem in order to keep those individuals in the analysis and to get sensible results? Best wishes Federico -- Federico C. F. Calboli Department of Epidemiology and Biostatistics Imperial College, St. Mary's Campus Norfolk Place, London W2 1PG Tel +44 (0)20 75941602 Fax +44 (0)20 75943193 f.calboli [.a.t] imperial.ac.uk f.calboli [.a.t] gmail.com
______________________________________________ R-help at r-project.org 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.
-- Dimitris Rizopoulos Assistant Professor Department of Biostatistics Erasmus University Medical Center Address: PO Box 2040, 3000 CA Rotterdam, the Netherlands Tel: +31/(0)10/7043478 Fax: +31/(0)10/7043014 Web: http://www.erasmusmc.nl/biostatistiek/
-- Federico C. F. Calboli Department of Epidemiology and Biostatistics Imperial College, St. Mary's Campus Norfolk Place, London W2 1PG Tel +44 (0)20 75941602 Fax +44 (0)20 75943193 f.calboli [.a.t] imperial.ac.uk f.calboli [.a.t] gmail.com
On 6/3/2011 12:32 PM, Federico Calboli wrote:
On 3 Jun 2011, at 11:27, Dimitris Rizopoulos wrote:
this sounds like a competing risks problem. Maybe you would be interested in a cause-specific hazard regression or the Fine& Gray model (http://cran.r-project.org/package=cmprsk).
I will look into that,but, biologically, parkinson leads to dementia (cause and effect rather than competing risk), and in fact all my subjects do have parkinson. Unless the two competing risks are 'risk of dementia' vs 'risk of death'.
Yes, this what I mean. Either you could be interested in the composite event dementia or death, whatever comes first, or you could be interested only in dementia, in which case death could be considered a competing risk. Best, Dimitris
Best F
Recently there was also a special issue in JSS on this topic (http://www.jstatsoft.org/v38). I hope it helps. Best, Dimitris On 6/3/2011 12:17 PM, Federico Calboli wrote:
I am writing to get a better handle on a warning I am getting from a coxph analysis I am doing. I am analysing age of onset of dementia *after* the onset of parkinson disease. My data looks like: age.park age.dem age.death censor x1 x2 x3 x4 1 76 87 88 0 16 33 E3 E3 2 75 84 84 0 33 36 E3 E3 3 77 81 81 1 NA NA<NA> <NA> 4 65 65 69 0 NA NA E4 E4 5 56 76 79 0 NA NA<NA> <NA> 6 62 72 72 1 NA NA<NA> <NA> ... Obviously some individuals (lines 1,2,5) will first develop parkinson, then a few years later, dementia. Some individuals will not develop dementia (lines 3 and 6, where age of death and age of dementia correspond, but the censor variable is 1). Some (more) unluky individuals develop parkinson and dementia at the same time (line 4). my coxph model looks like coxph(Surv(age.mot,age.dem, censor) ~ x1 + x2 + x3 + x4, mydata) and I get the warning: In Surv(age.mot, age.dem, censor) : Stop time must be> start time, NA created I am almost sure that this is due to the instances where age.park == age.dem, but there is nothing I can really do. So my question: how do I deal with the instances where age.park == age.dem in order to keep those individuals in the analysis and to get sensible results? Best wishes Federico -- Federico C. F. Calboli Department of Epidemiology and Biostatistics Imperial College, St. Mary's Campus Norfolk Place, London W2 1PG Tel +44 (0)20 75941602 Fax +44 (0)20 75943193 f.calboli [.a.t] imperial.ac.uk f.calboli [.a.t] gmail.com
______________________________________________ R-help at r-project.org 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.
-- Dimitris Rizopoulos Assistant Professor Department of Biostatistics Erasmus University Medical Center Address: PO Box 2040, 3000 CA Rotterdam, the Netherlands Tel: +31/(0)10/7043478 Fax: +31/(0)10/7043014 Web: http://www.erasmusmc.nl/biostatistiek/
-- Federico C. F. Calboli Department of Epidemiology and Biostatistics Imperial College, St. Mary's Campus Norfolk Place, London W2 1PG Tel +44 (0)20 75941602 Fax +44 (0)20 75943193 f.calboli [.a.t] imperial.ac.uk f.calboli [.a.t] gmail.com
Dimitris Rizopoulos Assistant Professor Department of Biostatistics Erasmus University Medical Center Address: PO Box 2040, 3000 CA Rotterdam, the Netherlands Tel: +31/(0)10/7043478 Fax: +31/(0)10/7043014 Web: http://www.erasmusmc.nl/biostatistiek/