Assessing Reliability and Validity in the Context of Planned Incomplete Data Structures for Multitrait-Multimethod Models

Brendan P. Bunting and Gary Adamson

This paper examines some designs for the effective use of the EM algorithm in minimising the burden of data collection in the context MTMM models. Within MTMM models it is possible to distinguish between reliability and validity, however, a serious limitation with this approach is that data are required on at least three occasions and with three different methods. A planned incomplete data structure is proposed which reduces substantially the amount of data required from each individual at a third point in time. The efficacy of this approach is tested via a series of simulations and the resulting parameter estimates are shown to be both precise and efficient.