Matched samples logistic regression in case-control studies with missing values : when to break the matches

• Simulated data sets are used to evaluate conditional and unconditionalmaximum likelihood estimation in an individual case-controldesign with continuous covariates when there are different ratesof excluded cases and different levels of other design parameters.The effectiveness of the estimation procedures is measured bymethod bias, variance of the estimators, root mean square error(RMSE) for logistic regression and the percentage of explainedvariation. Conditional estimation leads to higher RMSE thanunconditional estimation in the presence of missing observations,especially for 1:1 matching. The RMSE is higher for the smallerstratum size, especially for the 1:1 matching. The percentageof explained variation appears to be insensitive to missingdata, but is generally higher for the conditional estimationthan for the unconditional estimation. It is particularly goodfor the 1:2 matching design. For minimizing RMSE, a high matchingratio is recommended; in this case, conditional and unconditionallogistic regression models yield comparable levels of effectiveness.For maximizing the percentage of explained variation, the 1:2matching design with the conditional logistic regression modelis recommended.

Subject headings

NATURVETENSKAP  -- Matematik -- Sannolikhetsteori och statistik (hsv//swe)

NATURAL SCIENCES  -- Mathematics -- Probability Theory and Statistics (hsv//eng)

Keyword

Biostatistics

Logistic regression

Case-control studies

Missing values

Statistics, computer and systems science

Statistik, data- och systemvetenskap

Statistics

Statistik

Publication and Content Type

ref (subject category)

art (subject category)