Ensemble codes involving hippocampal neurons are at risk during delayed performance tests

Multielectrode recording techniques were used to record ensemble activity from 10 to 16 simultaneously active CA1 and CA3 neurons in the rat hippocampus during performance of a spatial delayed-nonmatch-to-sample task. Extracted sources of variance were used to assess the nature of two different types of errors that accounted for 30% of total trials. The two types of errors included ensemble “miscodes” of sample phase information and errors associated with delay-dependent corruption or disappearance of sample information at the time of the nonmatch response. Statistical assessment of trial sequences and associated “strength” of hippocampal ensemble codes revealed that miscoded error trials always followed delay-dependent error trials in which encoding was “weak,” indicating that the two types of errors were “linked.” It was determined that the occurrence of weakly encoded, delay-dependent error trials initiated an ensemble encoding “strategy” that increased the chances of being correct on the next trial and avoided the occurrence of further delay-dependent errors. Unexpectedly, the strategy involved “strongly” encoding response position information from the prior (delay-dependent) error trial and carrying it forward to the sample phase of the next trial. This produced a miscode type error on trials in which the “carried over” information obliterated encoding of the sample phase response on the next trial. Application of this strategy, irrespective of outcome, was sufficient to reorient the animal to the proper between trial sequence of response contingencies (nonmatch-to-sample) and boost performance to 73% correct on subsequent trials. The capacity for ensemble analyses of strength of information encoding combined with statistical assessment of trial sequences therefore provided unique insight into the “dynamic” nature of the role hippocampus plays in delay type memory tasks.

Influence analysis of robust Wald-type tests

We consider a robust version of the classical Wald test statistics for testing simple and composite null hypotheses for general parametric models. These test statistics are based on the minimum density power divergence estimators instead of the maximum likelihood estimators. An extensive study of their robustness properties is given though the influence functions as well as the chi-square inflation factors. It is theoretically established that the level and power of these robust tests are stable against outliers, whereas the classical Wald test breaks down. Some numerical examples confirm the validity of the theoretical results.

Generación automática de tests en DomJudge

Los jueces online están cobrando cada día más importancia, especialmente en el ámbito de la enseñanza. Su funciona-miento es simple, un juez/profesor sube un problema de programación con un enunciado y unos casos de prueba (entradas y salidas esperadas) al juez online. El alumno/concursante deberá subir el código que considera como solución al problema. Si el código del alumno devuelve las mismas salidas que las que se encuentran en los casos de prueba para las correspondientes entradas en los test-cases dada las mismas entradas, el código se considera correcto. En la Facultad de Informática de la Universidad Complutense de Madrid ha aparecido un juez virtual que cada vez está siendo más usado por los docentes como complemento a la hora de evaluar. Este juez online es DomJudge. DomJudge es un juez online desarrollado en la universidad de Utrecht con el fin de ser un juez virtual de concursos de programación. Su código es libre y se puede descargar y modificar, lo que lo hace ideal si se quiere adaptar su funcionamiento. Es bien sabido que generar casos de prueba de calidad es una tarea muy compleja. Existen diversas técnicas que ayudan a la generación automática de tests. Por ejemplo, la ejecución simbólica permite generar tests garantizando que todos los caminos de ejecución del programa son ejercitados hasta una cierta profundidad. El objetivo de este proyecto es hacer uso de estas técnicas de testing para generar casos de prueba de forma automática. Para que la autoevaluación realizada por estos jueces sea efectiva, los profesores deben proporcionar casos de prueba de calidad que son ejecutados automáticamente cuando los alumnos suben sus soluciones. Escribir estos casos de prueba resulta costoso y complejo, y es en este punto, donde el uso de jPET podría resultar muy útil. Los casos de prueba generados por jPET podrían servir como punto de partida a la hora de generar un conjunto de casos de prueba de calidad. Un aspecto muy interesante en este sentido, es que no es necesario que los programas de los alumnos uti licen el lenguaje Java. Lo único que sería necesario es que el profesor proporcione una solución escrita en Java. A partir de esta se podrían generar los casos de prueba iniciales. Si se diese el caso de que las soluciones de los alumnos viniesen escritas en Java, se podrían plantear enfoques más interesantes en los cuales los tests se forman a partir de ambas soluciones. La del alumno, para generarlos datos de entrada, y la del profesor, para chequear que las salidas para esas entradas son las correctas. Por eso hemos llevado a cabo no solo una traducción automática del xml generado por JPET en ficheros in y out. Sino una total integración con el sistema DomJudge para poder generar los casos de prueba sin intermediarios. Gracias a esta integración podemos facilitar el trabajo de los profesores a la hora de crear problemas en jueces online y conseguimos una generación automática de tests en DomJudge.

Portfolio efficiency tests with conditioning information using empirical likelihood estimation

We evaluate the use of Generalized Empirical Likelihood (GEL) estimators in portfolios efficiency tests for asset pricing models in the presence of conditional information. Estimators from GEL family presents some optimal statistical properties, such as robustness to misspecification and better properties in finite samples. Unlike GMM, the bias for GEL estimators do not increase as more moment conditions are included, which is expected in conditional efficiency analysis. We found some evidences that estimators from GEL class really performs differently in small samples, where efficiency tests using GEL generate lower estimates compared to tests using the standard approach with GMM. With Monte Carlo experiments we see that GEL has better performance when distortions are present in data, especially under heavy tails and Gaussian shocks.

Assessment of CO2 adsorption capacity on activated carbons by a combination of batch and dynamic tests

In this work, batch and dynamic adsorption tests are coupled for an accurate evaluation of CO2 adsorption performance for three different activated carbons obtained from olives stones by chemical activation followed by physical activation with CO2 at varying times, i.e. 20, 40 and 60 h. Kinetic and thermodynamic CO2 adsorption tests from simulated flue-gas at different temperature and CO2 pressure are carried out both in batch (a manometric equipment operating with pure CO2) and dynamic (a lab-scale fixed-bed column operating with CO2/N2 mixture) conditions. The textural characterization of the activated carbon samples shows a direct dependence of both micropore and ultramicropore volume on the activation time, hence AC60 has the higher contribution. The adsorption tests conducted at 273 and 293 K showed that, when CO2 pressure is lower than 0.3 bar, the lower the activation time the higher CO2 adsorption capacity and a ranking ωeq(AC20)>ωeq(AC40)>ωeq(AC60) can be exactly defined when T= 293 K. This result can be likely ascribed to a narrower pore size distribution of the AC20 sample, whose smaller pores are more effective for CO2 capture at higher temperature and lower CO2 pressure, the latter representing operating conditions of major interest for decarbonation of a flue-gas effluent. Moreover, the experimental results obtained from dynamic tests confirm the results derived from the batch tests in terms of CO2 adsorption capacity. It is important to highlight that the adsorption of N2 on the synthesized AC samples can be considered negligible. Finally, the importance of a proper analysis of characterization data and adsorption experimental results is highlighted for a correct assessment of CO2 removal performances of activated carbons at different CO2 pressure and operating temperature.