Partially Paramertic Estimation of Survival Function in the Right-censored Data

• Authors: Rossa, Agnieszka

Title variants

EN

Częściowo parametryczny estymator funkcji przeżycia dla danych prawostronnie cenzurowanych

Abstracts

EN

Abstract. In many medical, biological or economic follow-up studies the subject of observation is survival, failure or duration time, that is the length of time elapsed from a specific starting point to an event of interest. In engineering applications it may be the time to failure of piece of equipment, in medical trials - time to occurrence of a particular disease or time to death of a patient due to some specific disease, in economic studies - time of being unemployed and so on. In the analysis of survival-type variables one is often faced with right-censored observations. Sometimes it is impossible to measure the true failure time of an individual due to previous occurrence of some other event called competing event, which result in interruption of observation before the event occurs. It may be withdrawal of the subject from the study or failure from some causes other the one of interest or simply limitation on the length of study. If we are only interested in failure time, then the competing events can be regarded as right-censoring the event of interest. It means that for each individual we observe either the time to failure or the time to censoring and for censored individuals we know only that the time to failure is greater then the censoring time. In reliability studies censoring is often planned in order to obtain information sooner than it is otherwise possible. Instead of testing m units until they fail, the Type I censoring design is employed in which more then m units are tested but observation is terminated earlier at the end of some specified period x*. Those units, which failed before this time yield complete observations and the rest of them is right-censored. Despite such incompleteness of the data it is often desired to estimate survival function that is the probability P(X > x) that the true failure time X in the population of individuals exceeds x. The paper deals with a problem of estimating survival function in the right-censored data. Some improvements of the well-known Kaplan - Meier estimator are discussed and their properties are studied.

PL

W pracy omówione są dwa estymatory funkcji przeżycia, będące modyfikacją estymatora Kaplana-Meiera. Podstawowe własności statystyczne estymatorów zostały porównane za pomocą metod symulacyjnych.

Keywords

EN

censored data; survival function

• Publisher: Uniwersytet Łódzki. Wydawnictwo Uniwersytetu Łódzkiego
• Journal: Acta Universitatis Lodziensis. Folia Oeconomica
• Year: 2003
• Volume: 164

Dates

published

2003

Contributors

author

Rossa, Agnieszka

other

University of Łódź, Institute of Econometrics and Statistics

Identifiers

URI

http://hdl.handle.net/11089/6770