Cox Regression with Incomplete Covariate Measurements using the EM‐algorithm (Q4956049)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 1452464
Language Label Description Also known as
default for all languages
No label defined
    English
    Cox Regression with Incomplete Covariate Measurements using the EM‐algorithm
    scientific article; zbMATH DE number 1452464

      Statements

      Cox Regression with Incomplete Covariate Measurements using the EM‐algorithm (English)
      0 references
      0 references
      24 May 2000
      0 references
      Cox analysis
      0 references
      EM-algorithm
      0 references
      missing covariate data
      0 references
      survival analysis
      0 references
      The Cox regression model [\textit{D.R. Cox}, J. R. Stat. Soc., Ser. B 34, 187-220 (1972; Zbl 0243.62041)] is a very powerful tool in the analysis of censored failure data. The common practical way of handling missing covariate data in the Cox regression analysis is complete case analysis; that is, subjects with any missing covariate values are omitted from the analysis. Two other simple approaches are to use only those covariates that have complete measurements on every subject or to impute missing values. Both methods can lead to seriously biased estimates. The EM-algorithm, proposed by \textit{J.G. Ibrahim} [J. Am. Stat. Assoc. 85, 765-769 (1990)] is used to obtain maximum likelihood estimates of the regression parameters in generalized linear models with partially missing covariates. The technique was termed EM by the method of weights. In this paper the author proposes a new method which may be considered as a fully maximum likelihood approach. The Ibrahim technique is generalized to Cox regression analysis with missing values in the covariates. A fully model letting the unobserved covariate values be random and then maximizing the observed likelihood is specified. The asymptotic covariance matrix is estimated by the inverse information matrix. The missing data are allowed to be missing at random but also the non-ignorable non-response situation may in principle be considered. A Monte Carlo simulation is studied. The method is applied to data from a clinical trial studying the effect of predisposing vs placebo on survival with liver cirrhosis. Six covariates with three of them having missing values are considered.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references