Mechanism for Missing Data Incorporated in Joint Modelling of Ordinal Responses

Abstract

We analyse the problem of two clinically inseparable, repeatedly measured responses of ordinal type by also incorporating their missingness process. In our application these are the therapeutic effect and extent of side effects of fluvoxamine. In the case of a composite end point, the scientific questions addressed can be answered only when the responses are modelled jointly. As an extension of the methodology, several missingness not at random models were fitted to a set of observed data and shown to yield approximately the same result as their missingness at random counterparts, although it affects precision. In addition, the effect of various identifying restrictions on multiple imputation is investigated. An alternative numerical approximation method is suggested to reduce computational time.