Estimación de parámetros en modelos Rasch para la estimación de inseguridad alimentaria en una población.

Introduction: The Food Insecurity Experience Scale is a methodological proposal of the Voices of the Hungry Project of the FAO Statistics Division; is a set of questions whose objective is the measurement of food insecurity as a latent trait. This question module can be included in a survey, then with the response matrix, is possible to calculate prevalences of food insecurity through a Rasch model using the sum of the affirmative answers obtained by a person to the questions of the FIES module as a sufficient statistic for the estimation of the latent trait: Food insecurity. Materials and methods: A response matrix to the FIES module was simulated in R. Then a Rasch model was applied to this response matrix in order to obtain the theta and beta parameters according to different inferential frameworks: maximum likelihood, Bayesian, marginal and conditional, different R packages were used according to each estimation. Results and Discussion: The estimation of the parameters θ according to maximum likelihood and conditional estimation are not appropriate to estimate extreme scores. Conclusions: Bayesian and marginal estimation allow to estimate extreme scores for parameters θ, however, have high computational cost. The conditional estimate, which is the estimate currently used in the analytical protocol of the FIES scale at a global level, use pseudo scores to estimate the and and is the recommended estimation process if the severity of the latent trait in individuals/ households is the estimation required.