Bidimensional regression: Issues with interpolation
Thrash, T., Giannopoulos, I., & Schinazi, V.R. (2014). Bidimensional regression: Issues with interpolation. In P. Bello, M. Guarini, M. McShane, & B. Scassellati (Eds.), Proceedings of the 36th Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society.
We investigated the interpolation of missing values in data that were fit by bidimensional regression models. This addresses a problem in spatial cognition research in which sketch maps are used to assess the veracity of spatial representations. In several simulations, we compared samples of different sizes with different numbers of interpolated coordinate pairs. A genetic algorithm was used in order to estimate parameter values. We found that artificial inflation in the fit of bidimensional regression models increased with the percent of interpolated coordinate pairs. Furthermore, samples with fewer coordinate pairs resulted in more inflation than samples with more coordinate pairs. These results have important implications for statistical models, especially those applied to the analysis of spatial data.