La importancia del espacio geográfico para minimizar el error de muestras representativas

This paper discusses the importance of geographic space in the context of generating a sample framework for surveys, questioning the traditional statistical premise of randomness and independence of the number of observations. The contribution of quantitative geography in the generation of regionalization methodologies is analyzed, since these allow the improvement of the sampling error of the surveys, focusing mainly on urban areas, and in the presence of stratification variables with spatial autocorrelation. Regionalization algorithms with and without heuristic optimization processes are empirically tested, using census data, to subsequently define the level of error and establish comparisons against traditional random and two-stage random sampling, using a Monte Carlo procedure. The results obtained show a decrease of up to 20% in error against traditional methodologies or alternatively, a reduction of up to 100 cases with the same level of error. It is concluded that spatialized sampling methodologies with heuristic optimization offer advantages in urban areas, in the presence of spatial autocorrelation.