Estudio de la robustez del análisis discriminante multivariado : propuesta de un ajuste bootstrap para la significación estadística de los coeficientes de la técnica.

1. Realizar un estudio exhaustivo del Análisis Discriminante para evaluar su robustez con el fin de hacer las pertinentes recomendaciones al psicólogo aplicado; 2. Determinar criterios estadísticos que ayuden a las interpretaciones heurísticas de los coeficientes más relevantes, para la evaluación de las contribuciones de las variables a las funciones discriminantes. Primera investigación: Se trabajó con un diseño factorial 4x2x3x2x2 lo que supone 96 condiciones experimentales. Las cinco variables eran: a. Normalidad de las variables, b. Varianza de los grupos, c. Número de variables, d. Número de grupos, 5. Número de sujetos en cada grupo. Variable Dependiente: Para cada una de las 200 replicaciones Monte Carlo se obtuvieron las lambdas de Wilks, las V de Bartlett y su probabilidad asociada, como índice de la significación de criterio discriminante. Segunda investigación: Para esta investigación se replicó el diseño de la primera investigación, es decir, las 96 condiciones experimentales con todos los factores, otorgando ahora el perfil de diferencias grupales siguiente para las condiciones con tres grupos y para las condiciones con seis grupos. Se mantuvieron constantes las correlaciones entre las variables e iguales a las de la primera investigación, 0,70. El valor de los parámetros fue obtenido mediante el programa DISCRIMINANT del SPSS/PC+. Hardware: El trabajo de simulación se llevó a cabo con ocho ordenadores personales clónicos PC:PENTIUM/100 Mhz., con 16 MB de RAM. Software: Los procedimientos necesarios para la investigación fueron realizados en el lenguaje de programación GAUSS 386i, versión 3.1 (Aptech Systems, 1994). 1. Los métodos de simulación y concretamente, el método de muestreo bootstrap, son de gran utilidad para los estudios de robustez de las técnicas estadísticas, así como en los de inferencia estadística: cálculo de intervalos de confianza; 2. El Análisis Discriminante es una técnica robusta, siempre y cuando se cumpla la condición de homogeneidad de las varianzas; 3. El Análisis Discriminante no es robusto ante problemas de heterogeneidad en las siguientes condiciones: Con seis o menos variables,cuando los tamaños grupales son diferentes. Para tamaños iguales, si además se presenta una alteración conjunta de asimetría y apuntamiento; 4. Cuando la violación del supuesto de homogeneidad viene provocada porque la varianza mayor la presenta el grupo con menos sujetos la técnica se vuelve demasiado liberal, es decir, se produce un alto grado de error tipo I; 5. Los coeficientes de estructura son más estables e insesgados que los típicos; 6. Es posible determinar los intervalos confidenciales de los coeficientes de estructura mediante el procedimiento sugerido por Dalgleish (1994). 1. El Análisis Discriminante se puede utilizar siempre que se cumpla la condición de Homogeneidad de varianzas. Es por tanto, absolutamente necesario comprobar antes de realizar un Análisis Discriminante este principio, lo cual se puede llevar a cabo a través de cualquiera de los estadísticos pertinentes y, en especial, la prueba de Box; 2. Ante la heterogeneidad de varianzas si el número de variables independientes es seis o inferior, deberá tenerse en cuenta que el número de sujetos debe ser igual en todos los grupos y que las variables no presenten alteraciones conjuntas de asimetría y apuntamiento,por lo que, como paso previo deberá comprobarse la distribución de las variables y detectar si se presenta esta alteración. En cualquier otra condición, y ante la presencia de heterogeneidad de varianzas no se puede utilizar la técnica. Cuando el número de variables predictoras sea nueve o más, podrá utilizarse la técnica siempre, a excepción de diferentes tamaños grupales y no normalidad de las variables. El investigador aplicado deberá conocer la posibilidad que proponemos de apoyatura estadística para la toma de decisiones.

Aplicaci??n del m??todo Bootstrap al contraste de hip??tesis en la investigaci??n educativa

Resumen del autor. Resumen en castellano e ingl??s. Este art??culo se incluye en el monogr??fico 'Educaci??n social'

Multilevel bootstrap analysis with assumptions violated

Resumen tomado de la publicación

??Ha mejorado la productividad docente e investigadora de las universidades p??blicas espa??olas desde la aprobaci??n de la LOU? : evidencia a partir del bootstrap

Resumen basado en el de la publicaci??n

An application of the DEA double bootstrap to examine sources of efficiency in Bangladesh rice farming

In this article we examine sources of technical efficiency for rice farming in Bangladesh. The motivation for the analysis is the need to close the rice yield gap to enable food security. We employ the DEA double bootstrap of Simar and Wilson (2007) to estimate and explain technical efficiency. This technique overcomes severe limitations inherent in using the two-stage DEA approach commonly employed in the efficiency literature. From a policy perspective our results show that potential efficiency gains to reduce the yield gap are greater than previously found. Statistically positive influences on technical efficiency are education, extension and credit, with age being a negative influence.

Application of a double bootstrap to investigation of determinants of technical efficiency of farms in Central Europe

The paper provides one of the first applications of the double bootstrap procedure (Simar and Wilson 2007) in a two-stage estimation of the effect of environmental variables on non-parametric estimates of technical efficiency. This procedure enables consistent inference within models explaining efficiency scores, while simultaneously producing standard errors and confidence intervals for these efficiency scores. The application is to 88 livestock and 256 crop farms in the Czech Republic, split into individual and corporate.

Application of a double bootstrap to investigation of determinants of technical efficiency of farms in Central Europe

The paper provides one of the first applications of the double bootstrap procedure (Simar and Wilson 2007) in a two-stage estimation of the effect of environmental variables on non-parametric estimates of technical efficiency. This procedure enables consistent inference within models explaining efficiency scores, while simultaneously producing standard errors and confidence intervals for these efficiency scores. The application is to 88 livestock and 256 crop farms in the Czech Republic, split into individual and corporate.

Testing symmetry and homogeneity in the almost ideal demand system with co-integrated data using fully modified estimation and the bootstrap

Conventional seemingly unrelated estimation of the almost ideal demand system is shown to lead to small sample bias and distortions in the size of a Wald test for symmetry and homogeneity when the data are co-integrated. A fully modified estimator is developed in an attempt to remedy these problems. It is shown that this estimator reduces the small sample bias but fails to eliminate the size distortion.. Bootstrapping is shown to be ineffective as a method of removing small sample bias in both the conventional and fully modified estimators. Bootstrapping is effective, however, as a method of removing. size distortion and performs equally well in this respect with both estimators.

Bootstrap prediction intervals for autoregressive time series

The calculation of interval forecasts for highly persistent autoregressive (AR) time series based on the bootstrap is considered. Three methods are considered for countering the small-sample bias of least-squares estimation for processes which have roots close to the unit circle: a bootstrap bias-corrected OLS estimator; the use of the Roy–Fuller estimator in place of OLS; and the use of the Andrews–Chen estimator in place of OLS. All three methods of bias correction yield superior results to the bootstrap in the absence of bias correction. Of the three correction methods, the bootstrap prediction intervals based on the Roy–Fuller estimator are generally superior to the other two. The small-sample performance of bootstrap prediction intervals based on the Roy–Fuller estimator are investigated when the order of the AR model is unknown, and has to be determined using an information criterion.

Bootstrap percolation on homogeneous trees has 2 phase transitions

We study the threshold theta bootstrap percolation model on the homogeneous tree with degree b + 1, 2 <= theta <= b, and initial density p. It is known that there exists a nontrivial critical value for p, which we call p(f), such that a) for p > p(f), the final bootstrapped configuration is fully occupied for almost every initial configuration, and b) if p < p(f) , then for almost every initial configuration, the final bootstrapped configuration has density of occupied vertices less than 1. In this paper, we establish the existence of a distinct critical value for p, p(c), such that 0 < p(c) < p(f), with the following properties: 1) if p <= p(c), then for almost every initial configuration there is no infinite cluster of occupied vertices in the final bootstrapped configuration; 2) if p > p(c), then for almost every initial configuration there are infinite clusters of occupied vertices in the final bootstrapped configuration. Moreover, we show that 3) for p < p(c), the distribution of the occupied cluster size in the final bootstrapped configuration has an exponential tail; 4) at p = p(c), the expected occupied cluster size in the final bootstrapped configuration is infinite; 5) the probability of percolation of occupied vertices in the final bootstrapped configuration is continuous on [0, p(f)] and analytic on (p(c), p(f) ), admitting an analytic continuation from the right at p (c) and, only in the case theta = b, also from the left at p(f).

Is there a price puzzle in Brazil? An application of Bias-Corrected Bootstrap

Using data from the United States, Japan, Germany , United Kingdom and France, Sims (1992) found that positive innovations to shortterm interest rates led to sharp, persistent increases in the price level. The result was conÖrmed by other authors and, as a consequence of its non-expectable nature, was given the name "price puzzle" by Eichenbaum (1992). In this paper I investigate the existence of a price puzzle in Brazil using the same type of estimation and benchmark identiÖcation scheme employed by Christiano et al. (2000). In a methodological improvement over these studies, I qualify the results with the construction of bias-corrected bootstrap conÖdence intervals. Even though the data does show the existence of a statistically signiÖcant price puzzle in Brazil, it lasts for only one quarter and is quantitatively immaterial

Is there a price puzzle in Brazil? an application of bias- corrected bootstrap

Using data from the United States, Japan, Germany , United Kingdom and France, Sims (1992) found that positive innovations to shortterm interest rates led to sharp, persistent increases in the price leveI. The result was confirmed by other authors and, as a consequence of its non-expectable nature, was given the name "price puzzle" by Eichenbaum (1992). In this paper I investigate the existence of a price puzzle in Brazil using the same type of estimation and benchmark identification scheme employed by Christiano et aI. (2000). In a methodological improvement over these studies, I qualify the results with the construction of bias-corrected bootstrap confidence intervals. Even though the data does show the existence of a statistically significant price puzzle in Brazil, it lasts for .only one quarter and is quantitatively immaterial.