Diagnostics for Multiple Response Data
Surveys often allow arbitrary number of responses. Each of the categorical responses is referred as an item. Marginal modeling of items simultaneously requires to incorporate the dependence between items. We investigate deletion diagnostics as Cook distance and DBETA for these marginal models based on homogenous linear predictor (HLP) model fitting and compare results with the generalized estimation equations (GEE) approach.
Suesse, T. F. & Liu, I. (2008). Diagnostics for Multiple Response Data. In The Proceedings of PROBASTAT 2006, 2006, Smolenice Castle, Slovak Republic. Tatra Mountains Mathematical Publications, 39 105-113.