vol. 10 núm. 1 (2017)

Estimados lectores, es un gusto para mí retomar las labores como editor de la Revista Comunicaciones en Estadística. Quiero desear muchos éxitos en su labor a Hanwen Zhang editora anterior de la revista que a partir de este año se desempeña como decana de la facultad de Estadística de la Universidad. En esta ocasión, presento el primer número del volumen 10 de la revista Comunicaciones en Estadística. Puedo decir que ser editor de esta revista es una experiencia más que gratificante; encuentro en los mensajes de felicitación y de apoyo de nuestros lectores la motivaci´on de seguir trabajando fuertemente. Este número de la revista comienza con el trabajo de los investigadores V´ıctor M´arquez, Lelly Useche, Dulce Mesa y Ana Chac´on, quienes presentaron una metodolog´ıa de imputaci´on basada en ´arboles de regresión. Los autores encontraron expresiones de algunas propiedades del estimador propuesto, así como ventajas y desventajas del m´etodo, dando recomendaciones valiosas sobre el uso de este.

The estimate of the parameters of discrete probability distributions class (a,b) (Klugman et al. 2004, Escalante 2006) is presented in detail by methods of moments and maximum likelihood studied. A general algorithm is proposed to generate random numbers of distributions class (a,b). The results are presented so it can be implemented in any suitable programming language. Examples were made in R with real data taken from different disciplines.Keywords: Random models of frequency, estimation of distribution class (a,b), risk theory, the collective risk model, Panjer’s algorithm.  

We use the Multiple Factor Analysis (MFA) to built five classes of Latinamerican Universities from three known university rankings. These classes distinguish among universities with high level of specialization and low academic output, universities of excelence with low Scientific Leadership, universities with no good reputation and productive staff, universities with good reputation and few doctoral staff, and productive universities with high impact and low indicators of international collaboration. The factors produced by the MFA reveal some paradoxes corroborated in the classification by the fact that they counterpose the level of specialization vs. Productivity, the scientific leadership vs. impact and quality of the output, and they reveal too the possible inconvenience to include judging criteria, which result independent of leadership and impact.

Technological developments have made it possible for researchers in many areas to have large volumes of information for the same individual. Usually these data can be represented through curves or in general functions. From this arises a new field of study in statistics called Functional Data Analysis (FDA). In the FDA the basic unit of information is the complete function, rather than a set of values (Ramsay & Dalzell 1991). The usual statistical methods have been adapted to this situation, in particular the analysis of functional conglomerates by the k-means method has been developed. Since the brain activity responds to a wave function of the neuronal charge over time, the opportunity arises to apply the FDA to this type of record. The objective of this work is to describe the applicability of the functional cluster analysis by the k-means method to classify brain activity in Norvegicus Wistar rats. The conversion of the registers into wave functions was carried out using Fourier bases, which were analyzed according to the methodology developed in (Yamamoto 2012) and a simple correspondence analysis between the clusters and the phases of activity manually recorded in the hypnogram. The obtained conglomerates make a consistent unsupervised categorization, especially with respect to the attributes of frequency and regularity of the waves.

El problema de calibración no es reciente. Los trabajos en este tema fueron presentados inicialmente por Krutchkoff en la epoca de los 60's bajo un enfoque paramétrico y han sido ampliamente estudiados por otros autores desde diferentes enfoques. Las recientes investigaciones respecto al punto de cambio, han considerado supuestos adicionales y estimación usando modelos lineales mixtos. Se presenta una revisión exhaustiva de estos dos problemas y se puede observar que la vinculacion de estos no ha sido trabajado.  

In this work are compared using simulation methods for estimating Breslow, Efron and exact in Cox regression, to find the estimate of the model parameters, obtaining confidence intervals by resampling the Bootstrap, Jackknife and traditional asymptotic. Are generated samples of times using the inverse of the transformation, for models of exponential regression and Weibull. It illustrates the results of the amplitudes of the confidence intervals taking as a reference the regression estimate parametric. Showing the efficiency of these intervals.

This study analyses the evolution of the technical efficiency of the public universities in Colombia. The paper uses the information provided by the System of management Indicators developed by the government, related to 32 public universities during the period 2003-2012. The factorial scores of factor analysis are used to estimate the efficiency, education and research indexes used in the study. To estimation the education and research technical efficiency, a panel data stochastic production frontier model is used, controlling for the capacity index. The model assumes time-varying technical inefficiency. Finally, quantile regression is used to classify universities on quartiles 25, 50 and 75, based on their high, medium or low efficiency level. The results indicate that both education and research indexes increase with financial resources, physical capacity and, especially, professorial capacity.

In this paper, it is proposed the evaluation of a set of prior distributions for the scale parameters of the Zero-Inflated Poisson Regression model (ZIP). Traditionally the inverse-gamma distribution is used as prior for scale parameters. Some studies have shown that when the values of the hyperparameters of this distribution are very small, inferences are not adequate. We focus on evaluating three prior distributions for modeling scale parameters: inverse-gamma; half Cauchy and scaled beta 2 (SBeta2). The half Cauchy has been used in the situation in question and has proven to work properly. The SBeta2 is a heavy-tailed distribution that has better performance at the origin and at the right tail. A simulation study is developed, with which we intend to analyze the effect of the priordistributionassignedtothescaleparametersontheshrinkageoftheposterior estimates of parameters. Besides, the presence of outliers is evaluated regarding the adjustment of the corresponden values. This is done for each of the three prior distributions considering. The analysis focuses shrinkage of the posterior estimates of parameters and adjustment of outliers because the main criticisms on the use of the inverse-gamma distribution concentrate on this two issues. Finally an application is presented with real data.

 An imputation design is presented to combine classication and imputation in order to improve the quality of imputed datum. Imputation is done with completely randomized missing quantitative data and using regression trees. Media imputation techniques is compared, theoretical and empirically, using regression trees, in order to develop an integral classication and imputation strategy.Unbiased estimators were obtained developing the expected value of the estimator. Estimators proprieties were evaluated trough their variance and bias development, which showed non bias. as for the unbiased estimator variance of the media, suficiency was not proved for the media estimator.