Logistic modelling of longitudinal survey data with measurement error

A logistic model relating the rates of transition between two states to a vector of covariates is considered. Measurement error on the binary state variable can lead to severely biased parameter estimates. Estimation procedures which adjust for measurement error are proposed for different measurement models. Complex sampling designs are allowed for. The procedures are illustrated using data from the U.S. Panel Study of Income Dynamics, where the response is whether an individual is in a job with a union contract. It is found that adjusting for measurement error can be important.