Modelling general patterns of digit preference
In: del Castillo, J., Espinal, A., Puig, P. (Eds.): Proceedings of the 22nd International Workshop on Statistical Modelling, Barcelona, July 2-6, 2007, 148–153
Barcelona, Institut d´Estadistica de Catalunya, IDESCAT (2007)
Digit preference is a typical example of indirect observations of a latent distribution. A variation on the composite link model is a useful framework to uncover such latent distributions and it allows to estimate the proportions of counts that were transferred to neighbouring digits. We impose smoothness assumptions on the latent distribution since the estimating equations generally are singular or severely ill-conditioned. To estimate the misreported proportions we use a weighted least-squares regression with an added L1-penalty. The optimal smoothing parameters are found by minimizing the AIC. A simulation study and an actual application are presented.