The Generalized Least Squares Method (GMM) as a tool for causal analysis of spending, budget management and electoral results

Abstract

In the different fields of science, many times, there is a need to estimate the associations between variables, as an approach to understanding the interaction of one as a function of the others. It is usually done by applying restrictive models, such as analysis of variance and linear regression. This type of analysis requires that the dependent variable be continuous, have a normal and constant distribution of the mean and variance. However, when the dependent variable is discrete or categorical, the linear model is not viable. Faced with this impediment, the theory of linear models arises and is expanded to broader categories, which have been called Generalized Linear Models. This category assumes that all distribution functions are exponential, in which the normal distribution is located. In this sense, in this research, Generalized Least Squares methods were applied in their various variants: of moments, ordinary and feasible. These models allow calculating the parameters of models in which the dependent variable has a Poisson or multinomial distribution. In such a way that the Generalized Least Squares serve as a tool to analyze the effect of the elections on public spending and its relationship with the electoral results, analyzing the variables of a budgetary nature, derived from the possibility that the government in power continues or is re-elected. For this, data related to the states and municipalities of México in the period 2007 to 2019 are used.

Factores asociados a la eficiencia terminal en los posgrados en México: caso de estudio de dos programas de ciencias administrativas

“…Para poner a prueba las hipótesis establecidas, se utilizaron dos técnicas estadísticas: el análisis de componentes principales (ACP) y un modelo de regresión logística bivariada. En el ACP, se buscan las combinaciones lineales de predictores originales que reducen la cantidad de variables y explican una gran parte de la variación de un conjunto de datos (Nande et al, 2021).…”

“…Por su parte, en el modelo de regresión logística se busca la relación entre las dimensiones independientes: perfil académico (Ltec_o_pub y año_egre), rendimiento académico (Lexani, Prom_lic y Prom_maestría), perfil laboral (Niv_ocup y Lárea_prof) y sociodemográficas (Sexo y edad); y la dimensión dependiente: eficiencia terminal (se tituló sí o no) con la probabilidad de titularse. Los resultados se presentan como razones de momios, nombre técnico que se le da a la probabilidad de ocurrencia de un evento o variable (Nande et al, 2021).…”

“…Al cargar todos los indicadores en su respectiva dimensión, se asumió la validez del modelo en sus indicadores, los cuales tuvieron una confiabilidad aceptable conforme al índice alpha de Cronbach estipulado por Nande et al (2021). La siguiente figura expresa la relación entre las variables finales.…”