Continuous Ranked Probability Score Validation Methods in Mixture Bayesian Model for Microarray Data in Indonesia

Abstract

Abstract Validation in statistical modeling becomes a very important part to get information on how well the model has been built. Algorithm of Continuous Ranked Probability Score (CRPS) is a validation method of goodness of fit model in statistical modeling. A model that has a small CRPS value and has a small statistical significance, then the model is declared fit for data. Conversely, if a model has a large CRPS value, then the model is declared not fit for data. Several applications of the CRPS Algorithm have been developed for unimodal distribution models. Bayesian mixture is a modeling with Bayesian approach where data has a multimodal distribution. Characteristics of multimodal distribution are owned by microarray data in Indonesia, namely data on gene expression differences for several gene IDs from Chickpea plants in Indonesia. The purpose of this study was to obtain a performance from the Continuous Ranked Probability Score (CRPS) Algorithm as a goodness of fit model method in Bayesian Mixture Model (BMM) modeling for microarray data in Indonesia in a series of activities to find new varieties of Chiekpea plants that are resistant to attack by pathogenic fungal diseases Ascochyta Rabiei. The results of this study have succeeded in establishing the Algorithm of Continuous Ranked Probability Score (CRPS) for the distribution of normal mixture for data on gene expression differences of Chickpea plants in Indonesia as a result of microarray experiments with Bayesian approaches. BMM modeling on microarray data is declared fit because it has a small average value of CRPS, which is 0.0412 to 0.385.