Divide-and-Conquer solutions for estimating large consistent table sets
A new approach for the consistent estimation of a set coherent frequency tables
Numerical inconsistent results are often not tolerated within a set of coherent frequency tables. For example, if the number of “Groningers” who work as a “manager” is 50,000 according to one table, exactly the same number needs be published in an other table. Numerical consistent results are however not automatically achieved. Especially if several frequency tables need to be produced from multiple data sources, there is a risk of inconsistent results. Hence, there is clearly a need for estimation methods that lead to numerical consistent results.
Statistics Netherlands developed a Repeated Weighing (RW) method for the consistent estimation of a set of frequency tables. The scope of applicability of this method is however limited by several known estimation problems. The current paper proposes two alternative Divide-and-Conquer (D&C) methods that do not suffer from the estimation problems experienced with RW. For this reason the newly proposed methods can be much easier implemented than RW.