Överlevnadsanalys med konkurrerande dödsorsaker
(2015) STAH11 20142Department of Statistics
- Abstract
- In the field of statistics known as survival analysis, the time from a starting point to one or several endpoints is analysed. Traditional survival analysis deals with one endpoint, while competing risks theory addresses the situation with multiple endpoints. This thesis gives a theoretical introduction to both methods, emphasising the competing risks setting, which is of great importance in many fields of applied statistics, especially in biostatistics and epidemiology. Furthermore, two common regression methods for time-independent covariates, the cause-specific Cox model and the Fine and Gray subdistribution hazard model, are introduced. Their respective properties and efficiencies are analysed using both simulated data and a real... (More)
- In the field of statistics known as survival analysis, the time from a starting point to one or several endpoints is analysed. Traditional survival analysis deals with one endpoint, while competing risks theory addresses the situation with multiple endpoints. This thesis gives a theoretical introduction to both methods, emphasising the competing risks setting, which is of great importance in many fields of applied statistics, especially in biostatistics and epidemiology. Furthermore, two common regression methods for time-independent covariates, the cause-specific Cox model and the Fine and Gray subdistribution hazard model, are introduced. Their respective properties and efficiencies are analysed using both simulated data and a real dataset of breast cancer patients from southern Sweden. The two models are then assessed, with strengths and weaknesses discussed.
The results from the simulation study and the cancer data indicate that there are small differences between the models in most situations. However, the involved interpretability of the parameter estimates obtained from the Fine and Gray regression model could potentially cause problems in certain cases. Consequently, the Cox cause-specific model is the most appropriate one to use when evaluating the significance of parameters in a competing risks regression. (Less) - Popular Abstract (Swedish)
- Överlevnadsanalys är en gren av statistiken, där tiden från en startpunkt till dess att en eller flera händelser inträffar studeras. Klassisk överlevnadsanalys behandlar enbart en sluthändelse, medan teorin om konkurrerande risker möjliggör analys av multipla sluthändelser. Denna uppsats ger en teoretisk introduktion till båda metoderna men accentuerar modeller för analys av konkurrerande risker, vilka är mycket närvarande i många praktiska situationer, inte minst inom biostatistik och epidemiologi. Vidare beskrivs två vanliga regressionsmodeller vid analys av icke-tidsberoende kovariat, den orsaksspecifika Coxmodellen och Fine och Grays subdistributionshasardmodell. Deras respektive egenskaper och effektivitet analyseras med hjälp av... (More)
- Överlevnadsanalys är en gren av statistiken, där tiden från en startpunkt till dess att en eller flera händelser inträffar studeras. Klassisk överlevnadsanalys behandlar enbart en sluthändelse, medan teorin om konkurrerande risker möjliggör analys av multipla sluthändelser. Denna uppsats ger en teoretisk introduktion till båda metoderna men accentuerar modeller för analys av konkurrerande risker, vilka är mycket närvarande i många praktiska situationer, inte minst inom biostatistik och epidemiologi. Vidare beskrivs två vanliga regressionsmodeller vid analys av icke-tidsberoende kovariat, den orsaksspecifika Coxmodellen och Fine och Grays subdistributionshasardmodell. Deras respektive egenskaper och effektivitet analyseras med hjälp av simulerade data och verkliga data från sydsvenska bröstcancerpatienter. Därefter bedöms de båda modellernas styrkor och svagheter.
Resultaten från simuleringsstudien och från cancerdatan visar att det finns små skillnader mellan modellerna i de flesta situationer, men den praktiska tolkningen av de skattade parametrarna i Fine och Grays modell är mer invecklad än för Cox orsaksspecifika modell, varför slutsatsen blir att den sistnämnda modellen är att föredra vid regressionsanalys av kovariat vid närvaro av konkurrerande risker. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/5153495
- author
- Mehic, Adrian LU
- supervisor
-
- Jan Lanke LU
- organization
- course
- STAH11 20142
- year
- 2015
- type
- M2 - Bachelor Degree
- subject
- keywords
- Survival analysis, Epidemiology, Competing risks, Fine and Gray regression model, Cox proportional hazards model, Breast cancer
- language
- Swedish
- id
- 5153495
- date added to LUP
- 2015-03-12 12:07:25
- date last changed
- 2015-03-12 12:07:25
@misc{5153495, abstract = {{In the field of statistics known as survival analysis, the time from a starting point to one or several endpoints is analysed. Traditional survival analysis deals with one endpoint, while competing risks theory addresses the situation with multiple endpoints. This thesis gives a theoretical introduction to both methods, emphasising the competing risks setting, which is of great importance in many fields of applied statistics, especially in biostatistics and epidemiology. Furthermore, two common regression methods for time-independent covariates, the cause-specific Cox model and the Fine and Gray subdistribution hazard model, are introduced. Their respective properties and efficiencies are analysed using both simulated data and a real dataset of breast cancer patients from southern Sweden. The two models are then assessed, with strengths and weaknesses discussed. The results from the simulation study and the cancer data indicate that there are small differences between the models in most situations. However, the involved interpretability of the parameter estimates obtained from the Fine and Gray regression model could potentially cause problems in certain cases. Consequently, the Cox cause-specific model is the most appropriate one to use when evaluating the significance of parameters in a competing risks regression.}}, author = {{Mehic, Adrian}}, language = {{swe}}, note = {{Student Paper}}, title = {{Överlevnadsanalys med konkurrerande dödsorsaker}}, year = {{2015}}, }