National Statistics Institute

Products and Services / Publications / Working papers

imagen grupo productos y servicios
Portada documentos de trabajo

Two greedy algorithms for a binary quadratically constrained linear program in survey data editing

Doc: 03/2012

We propose a binary quadratically constrained linear program as an approach to selective editing. In a practice-oriented framework and allowing for some overediting whilst strictly fulfilling accuracy restrictions, we propose two greedy algorithms to find feasible suboptimal solutions. Their running times are quartic and cubic, respectively, in the number of sampling units and linear in the number of restrictions. We present computational evidence from several hundreds of instances randomly generated.

Palabras clave / Key words: Combinatorial optimization, quadratic constraint, linear program, greedy algorithm, selective editing.

Two greedy algorithms for a binary quadratically constrained linear program in survey data editing (Pdf 230 KB)
David Salgado, Ignacio Arbués, María Elisa Esteban

Important information about cookies

This website uses its own and third party cookies in order to provide a better service. These cookies are not used to collect personal information. If you continue browsing you are considered to accept their use.