PREDICTION FUNCTIONS FOR CATEGORICAL PANEL-DATA

Prediction of categorical responses in panel studies is considered. Pr ediction functions based on general conditional log-linear models are investigated for statistical properties both from a population perspec tive and a sampling perspective. Problems such as existence and unique ness of optimal prediction functions are addressed, and basic properti es of measures of prediction quality are examined. Estimation, consist ency and asymptotic normality are studied for the proposed parameter e stimates and measures of prediction quality.