Estimation of a probability in inverse binomial sampling under normalized linear-linear and inverse-linear loss

Sequential estimation of the success probability p in inverse binomial sampling is considered in this paper. For any estimator pˆ , its quality is measured by the risk associated with normalized loss functions of linear-linear or inverse-linear form. These functions are possibly asymmetric, with arbitrary slope parameters a and b for pˆ

p , respectively. Interest in these functions is motivated by their significance and potential uses, which are briefly discussed. Estimators are given for which the risk has an asymptotic value as p→0, and which guarantee that, for any p∈(0,1), the risk is lower than its asymptotic value. This allows selecting the required number of successes, r, to meet a prescribed quality irrespective of the unknown p. In addition, the proposed estimators are shown to be approximately minimax when a/b does not deviate too much from 1, and asymptotically minimax as r→∞ when a=b.

Estimation of a probability in inverse binomial sampling under normalized linear-linear and inverse-linear loss

Sequential estimation of the success probability $p$ in inverse binomial sampling is considered in this paper. For any estimator $\hatvap$, its quality is measured by the risk associated with normalized loss functions of linear-linear or inverse-linear form. These functions are possibly asymmetric, with arbitrary slope parameters $a$ and $b$ for $\hatvap < p$ and $\hatvap > p$ respectively. Interest in these functions is motivated by their significance and potential uses, which are briefly discussed. Estimators are given for which the risk has an asymptotic value as $p \rightarrow 0$, and which guarantee that, for any $p \in (0,1)$, the risk is lower than its asymptotic value. This allows selecting the required number of successes, $\nnum$, to meet a prescribed quality irrespective of the unknown $p$. In addition, the proposed estimators are shown to be approximately minimax when $a/b$ does not deviate too much from $1$, and asymptotically minimax as $\nnum \rightarrow \infty$ when $a=b$.

Estrategias de monitorización de CO2 y otros gases en los estudios de análogos naturales

La Universidad Politécnica de Madrid (UPM) y la Università degli Studi di Firenze (UniFi), bajo la coordinación técnica de AMPHOS21, participan desde 2009 en el proyecto de investigación “Estrategias de Monitorización de CO2 y otros gases en el estudio de Análogos Naturales”, financiado por la Fundación Ciudad de la Energía (CIUDEN) en el marco del Proyecto Compostilla OXYCFB300 (http://www.compostillaproject.eu), del Programa “European Energy Program for Recovery - EEPR”. El objetivo principal del proyecto fue el desarrollo y puesta a punto de metodologías de monitorización superficiales para su aplicación en el seguimiento y control de los emplazamientos donde se realice el almacenamiento geológico de CO2, analizando técnicas que permitan detectar y cuantificar las posibles fugas de CO2 a la atmósfera. Los trabajos se realizaron tanto en análogos naturales (españoles e italianos) como en la Planta de Desarrollo Tecnológico de Almacenamiento de CO2 de Hontomín. Las técnicas analizadas se centran en la medición de gases y aguas superficiales (de escorrentía y manantiales). En cuanto a la medición de gases se analizó el flujo de CO2 que emana desde el suelo a la atmósfera y la aplicabilidad de trazadores naturales (como el radón) para la detección e identificación de las fugas de CO2. En cuanto al análisis químico de las aguas se analizaron los datos geoquímicos e isotópicos y los gases disueltos en las aguas de los alrededores de la PDT de Hontomín, con objeto de determinar qué parámetros son los más apropiados para la detección de una posible migración del CO2 inyectado, o de la salmuera, a los ambientes superficiales. Las medidas de flujo de CO2 se realizaron con la técnica de la cámara de acúmulo. A pesar de ser una técnica desarrollada y aplicada en diferentes ámbitos científicos se estimó necesario adaptar un protocolo de medida y de análisis de datos a las características específicas de los proyectos de captura y almacenamiento de CO2 (CAC). Donde los flujos de CO2 esperados son bajos y en caso de producirse una fuga habrá que detectar pequeñas variaciones en los valores flujo con un “ruido” en la señal alto, debido a actividad biológica en el suelo. La medida de flujo de CO2 mediante la técnica de la cámara de acúmulo se puede realizar sin limpiar la superficie donde se coloca la cámara o limpiando y esperando al reequilibrio del flujo después de la distorsión al sistema. Sin embargo, los resultados obtenidos después de limpiar y esperar muestran menor dispersión, lo que nos indica que este procedimiento es el mejor para la monitorización de los complejos de almacenamiento geológico de CO2. El protocolo de medida resultante, utilizado para la obtención de la línea base de flujo de CO2 en Hontomín, sigue los siguiente pasos: a) con una espátula se prepara el punto de medición limpiando y retirando el recubrimiento vegetal o la primera capa compacta de suelo, b) se espera un tiempo para la realización de la medida de flujo, facilitando el reequilibrio del flujo del gas tras la alteración provocada en el suelo y c) se realiza la medida de flujo de CO2. Una vez realizada la medición de flujo de CO2, y detectada si existen zonas de anomalías, se debe estimar la cantidad de CO2 que se está escapando a la atmósfera (emanación total), con el objetivo de cuantificar la posible fuga. Existen un amplio rango de metodologías para realizar dicha estimación, siendo necesario entender cuáles son las más apropiadas para obtener el valor más representativo del sistema. En esta tesis se comparan seis técnicas estadísticas: media aritmética, estimador insegado de la media (aplicando la función de Sichel), remuestreo con reemplazamiento (bootstrap), separación en diferentes poblaciones mediante métodos gráficos y métodos basados en criterios de máxima verosimilitud, y la simulación Gaussiana secuencial. Para este análisis se realizaron ocho campañas de muestreo, tanto en la Planta de Desarrollo Tecnológico de Hontomón como en análogos naturales (italianos y españoles). Los resultados muestran que la simulación Gaussiana secuencial suele ser el método más preciso para realizar el cálculo, sin embargo, existen ocasiones donde otros métodos son más apropiados. Como consecuencia, se desarrolla un procedimiento de actuación para seleccionar el método que proporcione el mejor estimador. Este procedimiento consiste, en primer lugar, en realizar un análisis variográfico. Si existe una autocorrelación entre los datos, modelizada mediante el variograma, la mejor técnica para calcular la emanación total y su intervalo de confianza es la simulación Gaussiana secuencial (sGs). Si los datos son independientes se debe comprobar la distribución muestral, aplicando la media aritmética o el estimador insesgado de la media (Sichel) para datos normales o lognormales respectivamente. Cuando los datos no son normales o corresponden a una mezcla de poblaciones la mejor técnica de estimación es la de remuestreo con reemplazamiento (bootstrap). Siguiendo este procedimiento el máximo valor del intervalo de confianza estuvo en el orden del ±20/25%, con la mayoría de valores comprendidos entre ±3,5% y ±8%. La identificación de las diferentes poblaciones muestrales en los datos de flujo de CO2 puede ayudar a interpretar los resultados obtenidos, toda vez que esta distribución se ve afectada por la presencia de varios procesos geoquímicos como, por ejemplo, una fuente geológica o biológica del CO2. Así pues, este análisis puede ser una herramienta útil en el programa de monitorización, donde el principal objetivo es demostrar que no hay fugas desde el reservorio a la atmósfera y, si ocurren, detectarlas y cuantificarlas. Los resultados obtenidos muestran que el mejor proceso para realizar la separación de poblaciones está basado en criterios de máxima verosimilitud. Los procedimientos gráficos, aunque existen pautas para realizarlos, tienen un cierto grado de subjetividad en la interpretación de manera que los resultados son menos reproducibles. Durante el desarrollo de la tesis se analizó, en análogos naturales, la relación existente entre el CO2 y los isótopos del radón (222Rn y 220Rn), detectándose en todas las zonas de emisión de CO2 una relación positiva entre los valores de concentración de 222Rn en aire del suelo y el flujo de CO2. Comparando la concentración de 220Rn con el flujo de CO2 la relación no es tan clara, mientras que en algunos casos aumenta en otros se detecta una disminución, hecho que parece estar relacionado con la profundidad de origen del radón. Estos resultados confirmarían la posible aplicación de los isótopos del radón como trazadores del origen de los gases y su aplicación en la detección de fugas. Con respecto a la determinación de la línea base de flujo CO2 en la PDT de Hontomín, se realizaron mediciones con la cámara de acúmulo en las proximidades de los sondeos petrolíferos, perforados en los ochenta y denominados H-1, H-2, H-3 y H-4, en la zona donde se instalarán el sondeo de inyección (H-I) y el de monitorización (H-A) y en las proximidades de la falla sur. Desde noviembre de 2009 a abril de 2011 se realizaron siete campañas de muestreo, adquiriéndose más de 4.000 registros de flujo de CO2 con los que se determinó la línea base y su variación estacional. Los valores obtenidos fueron bajos (valores medios entre 5 y 13 g•m-2•d-1), detectándose pocos valores anómalos, principalmente en las proximidades del sondeo H-2. Sin embargo, estos valores no se pudieron asociar a una fuente profunda del CO2 y seguramente estuvieran más relacionados con procesos biológicos, como la respiración del suelo. No se detectaron valores anómalos cerca del sistema de fracturación (falla Ubierna), toda vez que en esta zona los valores de flujo son tan bajos como en el resto de puntos de muestreo. En este sentido, los valores de flujo de CO2 aparentemente están controlados por la actividad biológica, corroborado al obtenerse los menores valores durante los meses de otoño-invierno e ir aumentando en los periodos cálidos. Se calcularon dos grupos de valores de referencia, el primer grupo (UCL50) es 5 g•m-2•d-1 en las zonas no aradas en los meses de otoño-invierno y 3,5 y 12 g•m-2•d-1 en primavera-verano para zonas aradas y no aradas, respectivamente. El segundo grupo (UCL99) corresponde a 26 g•m-2•d- 1 durante los meses de otoño-invierno en las zonas no aradas y 34 y 42 g•m-2•d-1 para los meses de primavera-verano en zonas aradas y no aradas, respectivamente. Flujos mayores a estos valores de referencia podrían ser indicativos de una posible fuga durante la inyección y posterior a la misma. Los primeros datos geoquímicos e isotópicos de las aguas superficiales (de escorrentía y de manantiales) en el área de Hontomín–Huermeces fueron analizados. Los datos sugieren que las aguas estudiadas están relacionadas con aguas meteóricas con un circuito hidrogeológico superficial, caracterizadas por valores de TDS relativamente bajos (menor a 800 mg/L) y una fácie hidrogeoquímica de Ca2+(Mg2+)-HCO3 −. Algunas aguas de manantiales se caracterizan por concentraciones elevadas de NO3 − (concentraciones de hasta 123 mg/l), lo que sugiere una contaminación antropogénica. Se obtuvieron concentraciones anómalas de of Cl−, SO4 2−, As, B y Ba en dos manantiales cercanos a los sondeos petrolíferos y en el rio Ubierna, estos componentes son probablemente indicadores de una posible mezcla entre los acuíferos profundos y superficiales. El estudio de los gases disueltos en las aguas también evidencia el circuito superficial de las aguas. Estando, por lo general, dominado por la componente atmosférica (N2, O2 y Ar). Sin embargo, en algunos casos el gas predominante fue el CO2 (con concentraciones que llegan al 63% v/v), aunque los valores isotópicos del carbono (<-17,7 ‰) muestran que lo más probable es que esté relacionado con un origen biológico. Los datos geoquímicos e isotópicos de las aguas superficiales obtenidos en la zona de Hontomín se pueden considerar como el valor de fondo con el que comparar durante la fase operacional, la clausura y posterior a la clausura. En este sentido, la composición de los elementos mayoritarios y traza, la composición isotópica del carbono del CO2 disuelto y del TDIC (Carbono inorgánico disuelto) y algunos elementos traza se pueden considerar como parámetros adecuados para detectar la migración del CO2 a los ambientes superficiales. ABSTRACT Since 2009, a group made up of Universidad Politécnica de Madrid (UPM; Spain) and Università degli Studi Firenze (UniFi; Italy) has been taking part in a joint project called “Strategies for Monitoring CO2 and other Gases in Natural analogues”. The group was coordinated by AMPHOS XXI, a private company established in Barcelona. The Project was financially supported by Fundación Ciudad de la Energía (CIUDEN; Spain) as a part of the EC-funded OXYCFB300 project (European Energy Program for Recovery -EEPR-; www.compostillaproject.eu). The main objectives of the project were aimed to develop and optimize analytical methodologies to be applied at the surface to Monitor and Verify the feasibility of geologically stored carbon dioxide. These techniques were oriented to detect and quantify possible CO2 leakages to the atmosphere. Several investigations were made in natural analogues from Spain and Italy and in the Tecnchnological Development Plant for CO2 injection al Hontomín (Burgos, Spain). The studying techniques were mainly focused on the measurements of diffuse soil gases and surface and shallow waters. The soil-gas measurements included the determination of CO2 flux and the application to natural trace gases (e.g. radon) that may help to detect any CO2 leakage. As far as the water chemistry is concerned, geochemical and isotopic data related to surface and spring waters and dissolved gases in the area of the PDT of Hontomín were analyzed to determine the most suitable parameters to trace the migration of the injected CO2 into the near-surface environments. The accumulation chamber method was used to measure the diffuse emission of CO2 at the soil-atmosphere interface. Although this technique has widely been applied in different scientific areas, it was considered of the utmost importance to adapt the optimum methodology for measuring the CO2 soil flux and estimating the total CO2 output to the specific features of the site where CO2 is to be stored shortly. During the pre-injection phase CO2 fluxes are expected to be relatively low where in the intra- and post-injection phases, if leakages are to be occurring, small variation in CO2 flux might be detected when the CO2 “noise” is overcoming the biological activity of the soil (soil respiration). CO2 flux measurements by the accumulation chamber method could be performed without vegetation clearance or after vegetation clearance. However, the results obtained after clearance show less dispersion and this suggests that this procedure appears to be more suitable for monitoring CO2 Storage sites. The measurement protocol, applied for the determination of the CO2 flux baseline at Hontomín, has included the following steps: a) cleaning and removal of both the vegetal cover and top 2 cm of soil, b) waiting to reduce flux perturbation due to the soil removal and c) measuring the CO2 flux. Once completing the CO2 flux measurements and detected whether there were anomalies zones, the total CO2 output was estimated to quantify the amount of CO2 released to the atmosphere in each of the studied areas. There is a wide range of methodologies for the estimation of the CO2 output, which were applied to understand which one was the most representative. In this study six statistical methods are presented: arithmetic mean, minimum variances unbiased estimator, bootstrap resample, partitioning of data into different populations with a graphical and a maximum likelihood procedures, and sequential Gaussian simulation. Eight campaigns were carried out in the Hontomín CO2 Storage Technology Development Plant and in natural CO2 analogues. The results show that sequential Gaussian simulation is the most accurate method to estimate the total CO2 output and the confidential interval. Nevertheless, a variety of statistic methods were also used. As a consequence, an application procedure for selecting the most realistic method was developed. The first step to estimate the total emanation rate was the variogram analysis. If the relation among the data can be explained with the variogram, the best technique to calculate the total CO2 output and its confidence interval is the sequential Gaussian simulation method (sGs). If the data are independent, their distribution is to be analyzed. For normal and log-normal distribution the proper methods are the arithmetic mean and minimum variances unbiased estimator, respectively. If the data are not normal (log-normal) or are a mixture of different populations the best approach is the bootstrap resampling. According to these steps, the maximum confidence interval was about ±20/25%, with most of values between ±3.5% and ±8%. Partitioning of CO2 flux data into different populations may help to interpret the data as their distribution can be affected by different geochemical processes, e.g. geological or biological sources of CO2. Consequently, it may be an important tool in a monitoring CCS program, where the main goal is to demonstrate that there are not leakages from the reservoir to the atmosphere and, if occurring, to be able to detect and quantify it. Results show that the partitioning of populations is better performed by maximum likelihood criteria, since graphical procedures have a degree of subjectivity in the interpretation and results may not be reproducible. The relationship between CO2 flux and radon isotopes (222Rn and 220Rn) was studied in natural analogues. In all emissions zones, a positive relation between 222Rn and CO2 was observed. However, the relationship between activity of 220Rn and CO2 flux is not clear. In some cases the 220Rn activity indeed increased with the CO2 flux in other measurements a decrease was recognized. We can speculate that this effect was possibly related to the route (deep or shallow) of the radon source. These results may confirm the possible use of the radon isotopes as tracers for the gas origin and their application in the detection of leakages. With respect to the CO2 flux baseline at the TDP of Hontomín, soil flux measurements in the vicinity of oil boreholes, drilled in the eighties and named H-1 to H-4, and injection and monitoring wells were performed using an accumulation chamber. Seven surveys were carried out from November 2009 to summer 2011. More than 4,000 measurements were used to determine the baseline flux of CO2 and its seasonal variations. The measured values were relatively low (from 5 to 13 g•m-2•day-1) and few outliers were identified, mainly located close to the H-2 oil well. Nevertheless, these values cannot be associated to a deep source of CO2, being more likely related to biological processes, i.e. soil respiration. No anomalies were recognized close to the deep fault system (Ubierna Fault) detected by geophysical investigations. There, the CO2 flux is indeed as low as other measurement stations. CO2 fluxes appear to be controlled by the biological activity since the lowest values were recorded during autumn-winter seasons and they tend to increase in warm periods. Two reference CO2 flux values (UCL50 of 5 g•m-2•d-1 for non-ploughed areas in autumn-winter seasons and 3.5 and 12 g•m-2•d-1 for in ploughed and non-ploughed areas, respectively, in spring-summer time, and UCL99 of 26 g•m-2•d-1 for autumn-winter in not-ploughed areas and 34 and 42 g•m-2•d-1 for spring-summer in ploughed and not-ploughed areas, respectively, were calculated. Fluxes higher than these reference values could be indicative of possible leakage during the operational and post-closure stages of the storage project. The first geochemical and isotopic data related to surface and spring waters and dissolved gases in the area of Hontomín–Huermeces (Burgos, Spain) are presented and discussed. The chemical and features of the spring waters suggest that they are related to a shallow hydrogeological system as the concentration of the Total Dissolved Solids approaches 800 mg/L with a Ca2+(Mg2+)-HCO3 − composition, similar to that of the surface waters. Some spring waters are characterized by relatively high concentrations of NO3 − (up to 123 mg/L), unequivocally suggesting an anthropogenic source. Anomalous concentrations of Cl−, SO4 2−, As, B and Ba were measured in two springs, discharging a few hundred meters from the oil wells, and in the Rio Ubierna. These contents are possibly indicative of mixing processes between deep and shallow aquifers. The chemistry of the dissolved gases also evidences the shallow circuits of the Hontomín– Huermeces, mainly characterized by an atmospheric source as highlighted by the contents of N2, O2, Ar and their relative ratios. Nevertheless, significant concentrations (up to 63% by vol.) of isotopically negative CO2 (<−17.7‰ V-PDB) were found in some water samples, likely related to a biogenic source. The geochemical and isotopic data of the surface and spring waters in the surroundings of Hontomín can be considered as background values when intra- and post-injection monitoring programs will be carried out. In this respect, main and minor solutes, the isotopic carbon of dissolved CO2 and TDIC (Total Dissolved Inorganic Carbon) and selected trace elements can be considered as useful parameters to trace the migration of the injected CO2 into near-surface environments.

Interval-based ranking in noisy evolutionary multiobjective optimization

As one of the most competitive approaches to multi-objective optimization, evolutionary algorithms have been shown to obtain very good results for many realworld multi-objective problems. One of the issues that can affect the performance of these algorithms is the uncertainty in the quality of the solutions which is usually represented with the noise in the objective values. Therefore, handling noisy objectives in evolutionary multi-objective optimization algorithms becomes very important and is gaining more attention in recent years. In this paper we present ?-degree Pareto dominance relation for ordering the solutions in multi-objective optimization when the values of the objective functions are given as intervals. Based on this dominance relation, we propose an adaptation of the non-dominated sorting algorithm for ranking the solutions. This ranking method is then used in a standardmulti-objective evolutionary algorithm and a recently proposed novel multi-objective estimation of distribution algorithm based on joint variable-objective probabilistic modeling, and applied to a set of multi-objective problems with different levels of independent noise. The experimental results show that the use of the proposed method for solution ranking allows to approximate Pareto sets which are considerably better than those obtained when using the dominance probability-based ranking method, which is one of the main methods for noise handling in multi-objective optimization.

Extraterrestrial 3He estimation across the Paleocene-Eocene thermal maximum at ODP Site 208-1266

In the deep-sea, the Paleocene-Eocene Thermal Maximum (PETM) is often marked by clay-rich condensed intervals caused by dissolution of carbonate sediments, capped by a carbonate-rich interval. Constraining the duration of both the dissolution and subsequent cap-carbonate intervals is essential to computing marine carbon fluxes and thus testing hypotheses for the origin of this event. To this end, we provide new high-resolution helium isotope records spanning the Paleocene-Eocene boundary at ODP Site 1266 in the South Atlantic. The extraterrestrial 3He, 3HeET, concentrations replicate trends observed at ODP Site 690 by Farley and Eltgroth (2003, doi:10.1016/S0012-821X(03)00017-7). By assuming a constant flux of 3HeET we constrain relative changes in accumulation rates of sediment across the PETM and construct a new age model for the event. In this new chronology the zero carbonate layer represents 35 kyr, some of which reflects clay produced by dissolution of Paleocene (pre-PETM) sediments. Above this layer, carbonate concentrations increase for ~165 kyr and remain higher than in the latest Paleocene until 234 +48/-34 kyr above the base of the clay. The new chronology indicates that minimum d13C values persisted for a maximum of 134 +27/-19 kyr and the inflection point previously chosen to designate the end of the CIE recovery occurs at 217 +44/-31 kyr. This allocation of time differs from that of the cycle-based age model of Röhl et al. (2007, doi:10.1029/2007GC001784) in that it assigns more time to the clay layer followed by a more gradual recovery of carbonate-rich sedimentation. The new model also suggests a longer sustained d13C excursion followed by a more rapid recovery to pre-PETM d13C values. These differences have important implications for constraining the source(s) of carbon and mechanisms for its subsequent sequestration, favoring models that include a sustained release

Sentinel node biopsy for breast cancer: Using local results for estimation of risk to the patient

Background: Sentinel node biopsy (SNB) is being increasingly used but its place outside randomized trials has not yet been established. Methods: The first 114 sentinel node (SN) biopsies performed for breast cancer at the Princess Alexandra Hospital from March 1999 to June 2001 are presented. In 111 cases axillary dissection was also performed, allowing the accuracy of the technique to be assessed. A standard combination of preoperative lymphoscintigraphy, intraoperative gamma probe and injection of blue dye was used in most cases. Results are discussed in relation to the risk and potential consequences of understaging. Results: Where both probe and dye were used, the SN was identified in 90% of patients. A significant number of patients were treated in two stages and the technique was no less effective in patients who had SNB performed at a second operation after the primary tumour had already been removed. The interval from radioisotope injection to operation was very wide (between 2 and 22 h) and did not affect the outcome. Nodal metastases were present in 42 patients in whom an SN was found, and in 40 of these the SN was positive, giving a false negative rate of 4.8% (2/42), with the overall percentage of patients understaged being 2%. For this particular group as a whole, the increased risk of death due to systemic therapy being withheld as a consequence of understaging (if SNB alone had been employed) is estimated at less than 1/500. The risk for individuals will vary depending on other features of the particular primary tumour. Conclusion: For patients who elect to have the axilla staged using SNB alone, the risk and consequences of understaging need to be discussed. These risks can be estimated by allowing for the specific surgeon's false negative rate for the technique, and considering the likelihood of nodal metastases for a given tumour. There appears to be no disadvantage with performing SNB at a second operation after the primary tumour has already been removed. Clearly, for a large number of patients, SNB alone will be safe, but ideally participation in randomized trials should continue to be encouraged.

Enhanced detection-guided NLMS estimation of sparse FIR-modeled signal channels

In various signal-channel-estimation problems, the channel being estimated may be well approximated by a discrete finite impulse response (FIR) model with sparsely separated active or nonzero taps. A common approach to estimating such channels involves a discrete normalized least-mean-square (NLMS) adaptive FIR filter, every tap of which is adapted at each sample interval. Such an approach suffers from slow convergence rates and poor tracking when the required FIR filter is "long." Recently, NLMS-based algorithms have been proposed that employ least-squares-based structural detection techniques to exploit possible sparse channel structure and subsequently provide improved estimation performance. However, these algorithms perform poorly when there is a large dynamic range amongst the active taps. In this paper, we propose two modifications to the previous algorithms, which essentially remove this limitation. The modifications also significantly improve the applicability of the detection technique to structurally time varying channels. Importantly, for sparse channels, the computational cost of the newly proposed detection-guided NLMS estimator is only marginally greater than that of the standard NLMS estimator. Simulations demonstrate the favourable performance of the newly proposed algorithm. © 2006 IEEE.

Sequential, Bayesian geostatistics:a principled method for large data sets

The principled statistical application of Gaussian random field models used in geostatistics has historically been limited to data sets of a small size. This limitation is imposed by the requirement to store and invert the covariance matrix of all the samples to obtain a predictive distribution at unsampled locations, or to use likelihood-based covariance estimation. Various ad hoc approaches to solve this problem have been adopted, such as selecting a neighborhood region and/or a small number of observations to use in the kriging process, but these have no sound theoretical basis and it is unclear what information is being lost. In this article, we present a Bayesian method for estimating the posterior mean and covariance structures of a Gaussian random field using a sequential estimation algorithm. By imposing sparsity in a well-defined framework, the algorithm retains a subset of “basis vectors” that best represent the “true” posterior Gaussian random field model in the relative entropy sense. This allows a principled treatment of Gaussian random field models on very large data sets. The method is particularly appropriate when the Gaussian random field model is regarded as a latent variable model, which may be nonlinearly related to the observations. We show the application of the sequential, sparse Bayesian estimation in Gaussian random field models and discuss its merits and drawbacks.

Projected sequential Gaussian processes: flexible interpolation for large data sets

Recently within the machine learning and spatial statistics communities many papers have explored the potential of reduced rank representations of the covariance matrix, often referred to as projected or fixed rank approaches. In such methods the covariance function of the posterior process is represented by a reduced rank approximation which is chosen such that there is minimal information loss. In this paper a sequential framework for inference in such projected processes is presented, where the observations are considered one at a time. We introduce a C++ library for carrying out such projected, sequential estimation which adds several novel features. In particular we have incorporated the ability to use a generic observation operator, or sensor model, to permit data fusion. We can also cope with a range of observation error characteristics, including non-Gaussian observation errors. Inference for the variogram parameters is based on maximum likelihood estimation. We illustrate the projected sequential method in application to synthetic and real data sets. We discuss the software implementation and suggest possible future extensions.

Channel estimation and compensation in chromatic dispersion limited optical fast OFDM systems

We experimentally investigate the channel estimation and compensation in a chromatic dispersion (CD) limited 20Gbit/s optical fast orthogonal frequency division multiplexing (F-OFDM) system with up to 840km transmission. It is shown that symmetric extension based guard interval (GI) is required to enable CD compensation using one-tap equalizers. As few as one optical F-OFDM symbol with four and six pilot tones per symbol can achieve near-optimal channel estimation and compensation performance for 600km and 840km respectively.

Channel estimation and compensation in chromatic dispersion limited optical fast OFDM systems

We experimentally investigate the channel estimation and compensation in a chromatic dispersion (CD) limited 20Gbit/s optical fast orthogonal frequency division multiplexing (F-OFDM) system with up to 840km transmission. It is shown that symmetric extension based guard interval (GI) is required to enable CD compensation using one-tap equalizers. As few as one optical F-OFDM symbol with four and six pilot tones per symbol can achieve near-optimal channel estimation and compensation performance for 600km and 840km respectively.

Estimation of a Regression Function on a Point Process and its Application to Financial Ruin Risk Forecast

2000 Mathematics Subject Classification: Primary 60G55; secondary 60G25.

Projected sequential Gaussian processes:a C++ tool for interpolation of large datasets with heterogeneous noise

Heterogeneous datasets arise naturally in most applications due to the use of a variety of sensors and measuring platforms. Such datasets can be heterogeneous in terms of the error characteristics and sensor models. Treating such data is most naturally accomplished using a Bayesian or model-based geostatistical approach; however, such methods generally scale rather badly with the size of dataset, and require computationally expensive Monte Carlo based inference. Recently within the machine learning and spatial statistics communities many papers have explored the potential of reduced rank representations of the covariance matrix, often referred to as projected or fixed rank approaches. In such methods the covariance function of the posterior process is represented by a reduced rank approximation which is chosen such that there is minimal information loss. In this paper a sequential Bayesian framework for inference in such projected processes is presented. The observations are considered one at a time which avoids the need for high dimensional integrals typically required in a Bayesian approach. A C++ library, gptk, which is part of the INTAMAP web service, is introduced which implements projected, sequential estimation and adds several novel features. In particular the library includes the ability to use a generic observation operator, or sensor model, to permit data fusion. It is also possible to cope with a range of observation error characteristics, including non-Gaussian observation errors. Inference for the covariance parameters is explored, including the impact of the projected process approximation on likelihood profiles. We illustrate the projected sequential method in application to synthetic and real datasets. Limitations and extensions are discussed. © 2010 Elsevier Ltd.

The difference model with guessing explains interval bias in two-alternative forced- chocice detection procedures

The standard difference model of two-alternative forced-choice (2AFC) tasks implies that performance should be the same when the target is presented in the first or the second interval. Empirical data often show “interval bias” in that percentage correct differs significantly when the signal is presented in the first or the second interval. We present an extension of the standard difference model that accounts for interval bias by incorporating an indifference zone around the null value of the decision variable. Analytical predictions are derived which reveal how interval bias may occur when data generated by the guessing model are analyzed as prescribed by the standard difference model. Parameter estimation methods and goodness-of-fit testing approaches for the guessing model are also developed and presented. A simulation study is included whose results show that the parameters of the guessing model can be estimated accurately. Finally, the guessing model is tested empirically in a 2AFC detection procedure in which guesses were explicitly recorded. The results support the guessing model and indicate that interval bias is not observed when guesses are separated out.

Cervical cancer precursors and hormonal contraceptive use in HIV-positive women: application of a causal model and semi-parametric estimation methods.

OBJECTIVE: To demonstrate the application of causal inference methods to observational data in the obstetrics and gynecology field, particularly causal modeling and semi-parametric estimation. BACKGROUND: Human immunodeficiency virus (HIV)-positive women are at increased risk for cervical cancer and its treatable precursors. Determining whether potential risk factors such as hormonal contraception are true causes is critical for informing public health strategies as longevity increases among HIV-positive women in developing countries. METHODS: We developed a causal model of the factors related to combined oral contraceptive (COC) use and cervical intraepithelial neoplasia 2 or greater (CIN2+) and modified the model to fit the observed data, drawn from women in a cervical cancer screening program at HIV clinics in Kenya. Assumptions required for substantiation of a causal relationship were assessed. We estimated the population-level association using semi-parametric methods: g-computation, inverse probability of treatment weighting, and targeted maximum likelihood estimation. RESULTS: We identified 2 plausible causal paths from COC use to CIN2+: via HPV infection and via increased disease progression. Study data enabled estimation of the latter only with strong assumptions of no unmeasured confounding. Of 2,519 women under 50 screened per protocol, 219 (8.7%) were diagnosed with CIN2+. Marginal modeling suggested a 2.9% (95% confidence interval 0.1%, 6.9%) increase in prevalence of CIN2+ if all women under 50 were exposed to COC; the significance of this association was sensitive to method of estimation and exposure misclassification. CONCLUSION: Use of causal modeling enabled clear representation of the causal relationship of interest and the assumptions required to estimate that relationship from the observed data. Semi-parametric estimation methods provided flexibility and reduced reliance on correct model form. Although selected results suggest an increased prevalence of CIN2+ associated with COC, evidence is insufficient to conclude causality. Priority areas for future studies to better satisfy causal criteria are identified.