Search

El ANOVA es ampliamente usado en diversos campos de la investigaci on, es por ello que veri car el cumplimiento de los supuestos es muy importante; especialmente el de homogeneidad de varianzas. Actualmente existen una buena cantidad de pruebas para este ultimo, sin embargo se tiene una limitante en el ambito pr actico cuando la muestra es censurada. El objetivo de este trabajo fue investigar el comportamiento de la prueba de Bartlett bajo censura por la derecha del tipo II. Se utiliza el m etodo de m axima verosimilitud para estimar las varianzas, se investiga el comportamiento asint otico (m etodo Delta) del estad stico de prueba bajo censura y por medio del m etodo Monte Carlo a 10,000 repeticiones (donde tanto el tama~no y censura de n1 sea = o 6= a n2), para obtener la distribuci on del estad stico de Bartlett bajo censura. Encontramos que contar con esta metodolog a es util ya que conforme aumenta la censura en las muestras, la media y varianza de la distribuci on del estad stico de Bartlett aumenta considerablemente. Las tablas de valores de la distribuci on fueron generadas as como el tama~no, potencia y robustez de la prueba. Se gener o tambi en la metodolog a para la prueba de Levene para muestras censuradas por la derecha tipo I. Finalmente se ejempli ca el uso de ambas pruebas en casos de reproducci on ovina. _______________ A VARIANCE HOMOGENEITY TEST IN CENSORED NORMAL SAMPLES. ABSTRACT: A variance homogeneity test for right censored type II sample is proposed. The test is based on Bartlett's statistic. By using simulation the critical values of the null distribution of Bartlett's statistic for testing homogeneity of variances of two normal populations is obtained when the sample sizes and censoring levels are not equal. Also, we investigated the properties of the proposed test (size, power and robustness). The asymptotic distribution of the test statistic is also obtained. Results shows that the distribution of the test statistic distribution depends on censure level. The test keeps the size at di erent signi cance levels,the test is consistent. Furthemore, a metodology for Levene's test under type I right censure is proposed. Finally, some example of application in animal reproduction research is also given for both Bartlett's and Levene's test under censure.

Objective: To identify the factors that influence the use of dental services in 4-7-year-olds and in 10-13-year-olds resident in the cities of Talca (Chile) and Montreal (Canada). Methods: A nonprobabilistic cross-sectional study was carried out in 147 boys and girls in Talca and in 94 boys and girls in Montreal between 2009 and 2011. Sociodemographic variables were recorded in parents and children, including age and sex. Data were also gathered on parental education, family composition, and proximity to health centers within neighborhoods. The data were analyzed with Fisher's exact test and the robust Cox regression model (with constant time) with a significance level of 0,05. Results: In Talca, parental education was significantly associated with dental care visits at least twice a year. The children of parents with university education were 2.20 times more likely to consult a dentist (95% CI: 1.30-3.73). Children whose parents perceived their children's health positively were 53% (OR = 0,47; 95% Cl: 0,28-0,77) less likely to consult a dentist. In Montreal, the children of parents with university education were 2.10 times more likely to consult a dentist (95% CI: 1.17-3.76), while older children (10-13 years) were 2.11 (95% CI: 1.15-3.88) times more likely to consult a dentist. Conclusions: In both cities, parental education level was associated with the use of dental services. Refereed/Peer-reviewed