Linking Errors between Two Populations and Tests: A Case Study in International Surveys in Education