Pseudo-observations in survival analysis
Translated title
Pseudo-observationer i overlevelses analyse
Author
Mortensen, Rikke Nørmark
Term
4. term
Education
Publication year
2013
Submitted on
2013-06-03
Pages
78
Abstract
Overlevelsesdata beskriver tiden, der går, før en begivenhed indtræffer, for eksempel død eller sygdomstilbagefald. I denne type data oplever man ofte censurering: For nogle personer nås begivenheden ikke, før undersøgelsen slutter, eller de forlader undersøgelsen, så deres fulde forløb er ukendt. Det gør, at almindelige statistiske metoder ikke altid er egnede. Derfor bruges særlige metoder som Cox’ proportional hazards-model, en regressionsmodel der kobler forklarende variable (kovariater) til den øjeblikkelige risiko for en begivenhed (også kaldet farefunktionen). Dette projekt undersøger en metode baseret på pseudo-observationer fra jackknife-teori. Ideen er at konstruere stedfortrædende værdier for hver deltager, så man kan analysere overlevelsesdata med standard regressions- og andre statistiske metoder, selv når nogle forløb er censurerede. Pseudo-observationer adresserer dermed udfordringen med, at der ikke findes fuldstændige observationer for alle individer. Projektet vurderer potentialet og effektiviteten af pseudo-observationer til regressionsanalyse af overlevelsesdata og sammenligner denne tilgang med den traditionelle Cox-model.
Survival data track the time until an event occurs, such as death or disease relapse. These datasets often include censoring: for some people, the event is not observed before the study ends or they leave the study, so their full outcome is unknown. Because of this, standard statistical methods are not always suitable. Specialized methods are therefore used, notably the Cox proportional hazards model, a regression approach that links explanatory variables (covariates) to the instantaneous risk of the event (the hazard). This project examines a method based on pseudo-observations from jackknife theory. The idea is to create proxy values for each participant so that survival data can be analyzed with standard regression and other common statistical tools, even when some outcomes are censored. In this way, pseudo-observations address the challenge of not having complete observations for everyone. The project assesses the potential and efficiency of pseudo-observations for regression analyses of survival data and compares this approach with the traditional Cox model.
[This abstract was generated with the help of AI]
Documents