A two-piece scale mixture normal measurement error models for replicated data

‎We develop a new class of flexible replicated measurement error models (RMEM) based on the normal two-piece scale mixture (TP-SMN) family to model the distribution of the latent variable‎. ‎In the proposed approach‎, ‎the replicated observations are jointly modeled by a mixture of two components from a scale mixture skew-normal (SMSN) density‎. ‎The flexibility of this class can enable the simultaneous accommodation of skewness‎, ‎outliers‎, ‎and multimodality‎. ‎The proposed connection between the unobserved covariates and the response facilitates the construction of an EM-type algorithm to perform maximum likelihood estimation‎. ‎The effectiveness of the maximum likelihood estimations is studied through the simulation studies‎. ‎Also‎, ‎the method is applied to analyze continuing survey data on food intake by individuals on diet habits‎.

کلیدواژه ها:

ECM algorithm ، Equation error ، Replicated measurement error model ، Two-piece scale mixture normal