Model-based incomplete data analysis with an application to occupational mobility and migration accounts

In many planning and policy research settings available secondary data sources may be incapable of answering pertinent research questions because certain variable combinations are unavailable. One solution to this constraint is to try to construct the desired data using information from multiple data sources and prior information. Current methods for accomplishing this task tend to focus predominantly on updating transaction matrices (input-output tables, transportation flows, or interregional migration accounts) and emphasize an algorithmic approach to the problem. This paper attempts to broaden the applications and generalize the solution by extending the model-based approach to incomplete data analysis advocated by Willekens (1982). The log-linear model is presented here as a flexible platform for incomplete data analysis and a path diagram describes several alternative modeling approaches; different paths are determined by the level of available information. The paper concludes with an application to incomplete occupational migration and mobility tables.