Bias in weighted likelihood estimation when using the Rasch model

There are three methods of maximum likelihood estimation that are commonly used with Rasch item response models: Joint maximum likelihood, which provides point estimates for both person and item parameters; conditional maximum likelihood, which produces estimates for the item parameters only, and marginal maximum likelihood, which produces estimates of the item parameters and parameters of the distribution from which it has been assumed that the persons have been sampled. A number of approaches have been suggested for obtaining ability estimates for subsequent use in reporting or analysis when either conditional or marginal methods have been used to obtain the item parameter estimates.