Estimación máximo verosímil y mínimos cuadrados en el modelo estadístico lineal mixto: Metodología y análisis
The present study aims to establish the estimation methodology and analysis in the mixed linear model, these models are mainly used to describe the relationship between a response variable and some covariates in the data that are grouped according to one or more factors of classification, where the...
Salvato in:
| Pubblicazione: |
2014
|
|---|---|
| Soggetti: | |
| Accesso online: | http://hdl.handle.net/10872/14810 |
| Tags: |
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| Riassunto: | The present study aims to establish the estimation methodology and analysis in the mixed linear model, these models are mainly used to describe the relationship between a response variable and some covariates in the data that are grouped according to one or more factors of classification, where the same are fixed and random effects, in addition provide the possibility of analyzing data no dependent, unbalanced and lacking of normality. For the present investigation described the theoretical procedures of Maximum Likelihood, Restricted Maximum Likelihood (ML and REML) and Least Squares (GLM) estimation methods. In order to determine which of the estimates of the model would be most suitable, was took into consideration the mean square residual (MSE) and then the best method will be that resulting with the lower (MSE), since this is considered as a useful tool to decide what so proper is a model and determine the extent in which the model does not conform to the information, or remove certain terms can be simplified in ways which promote model. It provides a way to choose the best estimator: a minimum mean square residual often indicates a minimum variation, and is therefore considered a good estimator. To contrast the estimation methods previously mentioned, were performed two examples one for simulated data (balanced and unbalanced) normally distributed with homogeneous variances and other with actual unbalanced data focused on the area of genetics, especially in cattle. Additionally different models at the same set of real data were adjusted, so it was necessary to select the most suitable for them, therefore the information criteria were used as: Akaike Information Criterion (AIC) and Schwarz Information Criterion or Bayesian (BIC). It was obtained the maximum likelihood as the method of estimation with the lower square medium residual. |
|---|