Abstract
Data with measurement error in the outcome, covariates, or both are not uncommon, particularly with the increased use of routinely collected data for biomedical research. With error-prone data, often only a subsample of study data is validated; such settings are known as two-phase studies. The sieve maximum likelihood estimator (SMLE), which combines the error-prone data on all records with the validated data on a subsample, is a highly efficient and robust method to analyze such data. However, given their complexity, a computationally efficient and user-friendly tool is needed to obtain the SMLEs. The R package sleev fills this gap by making semiparametric likelihood-based inference using the SMLEs for error-prone two-phase data in settings with binary and continuous outcomes. Functions from this package can be used to analyze data with error-prone binary or continuous responses and error-prone covariates.