Recursive least—squares approach to data transferability: exposition and numerical results

Abstract

Data transferability refers to the transfer of information from a set of independently estimable models to a new but structurally equivalent model for which data on some predictors are missing.Using a random-coefficient regression framework, this paper discusses a recursive least-squares approach to execute the transfer and estimator all parameters of the new model.Numerical results document the method's sensitivity to practically relevant dimension and, as such, as establish its general applicability.

Keywords:

• Transferability
• Random coefficient regression
• Recursive least—Squares

^*Assistant Professor and Professor of Marketing, INSEAD, Fontainebleau, France.Funding was provided by the R&D Department of INSEAD.This research benefited from critical comments by colleagues at Columbia University and INSEAD

^*Assistant Professor and Professor of Marketing, INSEAD, Fontainebleau, France.Funding was provided by the R&D Department of INSEAD.This research benefited from critical comments by colleagues at Columbia University and INSEAD

Notes

^*Assistant Professor and Professor of Marketing, INSEAD, Fontainebleau, France.Funding was provided by the R&D Department of INSEAD.This research benefited from critical comments by colleagues at Columbia University and INSEAD