Pasar al contenido principal

NUEVOS AVANCES METODOLOGICOS Y COMPUTATIONALES EN ESTADISTICA NO PARAMETRICA Y SEMIPARAMETRICA

PID2020-118101GB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2020
Plan Estatal 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDAD DE VIGO

Resultados relacionados

Resultados totales (Incluyendo duplicados): 19
Encontrada(s) 1 página(s)

Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Estimation of bankruptcy rules with a priori unions for establishing new systems of quotas

  • Saavedra Nieves, Alejandro
  • Saavedra Nieves, Paula
This paper addresses a sampling procedure for estimating extensions of the random arrival rule to those bankruptcy situations where there exist a priori unions. It is based on simple random sampling with replacement and it adapts an estimation method of the Owen value for transferable utility games with a priori unions, especially useful when the set of involved agents is sufficiently large. We analyse the theoretical statistical properties of the resulting estimator as well as we provide some bounds for the incurred error. Its performance is evaluated on two well-studied examples in literature where this allocation rule can be exactly obtained. Finally, we apply this sampling method to provide a new quota system for the milk market in Galicia (Spain) for checking the role of different territorial structures when they are taken as a priori unions. The resulting quotas estimator is also compared with two classical rules in bankruptcy literature.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

HDiR: An R Package for Computation and Nonparametric Plug-in Estimation of Directional Highest Density Regions and General Level Sets

  • Saavedra Nieves, Paula
  • Crujeiras Casais, Rosa María
A deeper understanding of a distribution support, being able to determine regions of a certain (possibly high) probability content is an important task in several research fields. Package HDiR for R is designed for exact computation of directional (circular and spherical) highest density regions and density level sets when the density is fully known. Otherwise, HDiR implements nonparametric plug-in methods based on different kernel density estimates for reconstructing this kind of sets. Additionally, it also allows the computation and plug-in estimation of level sets for general functions (not necessarily a density). Some exploratory tools, such as suitably adapted distances and scatterplots, are also implemented. Two original datasets and spherical density models are used for illustrating HDiR functionalities
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Caspase 3 and Cleaved Caspase 3 Expression in Tumorogenesis and Its Correlations with Prognosis in Head and Neck Cancer: A Systematic Review and Meta-Analysis

  • Silva, Fábio França Vieira e
  • Padín Iruegas, María Elena
  • Caponio, Vito Carlo Alberto
  • Lorenzo Pouso, Alejandro Ismael
  • Saavedra Nieves, Paula
  • Chamorro Petronacci, Cintia Micaela
  • Suárez Peñaranda, José Manuel
  • Pérez-Sayáns García, Mario
Head and neck cancer (HNC) is an ascending and agressive disease. The search for new molecular markers is emerging to solve difficulties in diagnosis, risk management, prognosis and effectiveness of treatments. Proteins related to apoptotic machinery have been identified as potential biomarkers. Caspase 3 is the main effector caspase and has a key role in apoptosis. The objective of this systematic review and meta-analysis is to review studies that analyze changes in Caspase 3 and Cleaved Caspase 3 expression both in oral premalignant disorders (OPMD) as well as in head and neck cancer (HNC). This study also proposes to review the prognostic values associated with HNC according to the expression of Caspase 3. Medline (via PubMed), EMBASE, Scopus, Cochrane, Web of Science and Grey Literature Database were screened from inception to june of 2022 and 18 studies were selected and 8 were included in the prognostic meta-analysis. Results related to the comparison of Caspase 3 expression demonstrated similar expression of Caspase 3 in HNC, with an average of 51.9% (9.5–98.1) showing high/moderate expression compared to 45.7% (14.6–84.7) in OPMD. Of interest, Cleaved Caspase 3 resulted incresed in HNC when compared with OPMD, being 73.3% (38.6–88.3) versus 22.9% (7.1–38.7). Pooled Fixed effect of HR values (95% CI) for OS related to Caspase 3 IHC expression in HNC patients was 1.48 (95% CI 0.95–2.28); also, the rate of heterogeneity was low, as revealed by I2 = 31%. For DFS was 1.07 (95% CI 0.79–1.45) with I2 = 0% and DSS showed a HR of 0.88 (95% CI 0.69–1.12) with I2 = 37%. Caspase 3 and Cleaved Caspase 3 expression could be linked with malignancy progression, but the expression of Caspase 3 did not influence the prognosis of patients with HNC.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

On the estimation of the core for TU games

  • Saavedra Nieves, Alejandro
  • Saavedra Nieves, Paula
The core of a transferable utility (TU) game, if it is not empty, is prescribed by the set of all stable allocations. The exact determination of the core reaches exponential time complexity. Therefore, its exact computation is often avoided as the number of players increases. In this work, we propose an estimator for the core of a TU game based on the statistical theory of set estimation. Concretely, we provide a core reconstruction that is obtained in polynomial time for general dimension. Additionally, convergence rates for the estimation error are derived. Finally, a consistent core-center estimator is established as a geometrical application of this methodology.
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

seq2R: an R package to detect change points in DNA sequences

  • Martinez Villanueva, Nora
  • Sestelo Perez, Marta
  • Fonseca, Miguel M.
  • Roca Pardiñas, Javier
Identifying the mutational processes that shape the nucleotide composition of the mitochondrial genome (mtDNA) is fundamental to better understand how these genomes evolve. Several methods have been proposed to analyze DNA sequence nucleotide composition and skewness, but most of them lack any measurement of statistical support or were not developed taking into account the specificities of mitochondrial genomes. A new methodology is presented, which is specifically developed for mtDNA to detect compositional changes or asymmetries (AT and CG skews) based on nonparametric regression models and their derivatives. The proposed method also includes the construction of confidence intervals, which are built using bootstrap techniques. This paper introduces an R package, known as seq2R, that implements the proposed methodology. Moreover, an illustration of the use of seq2R is provided using real data, specifically two publicly available complete mtDNAs: the human (Homo sapiens) sequence and a nematode (Radopholus similis) mitogenome sequence., Ministerio de Ciencia e Innovación | Ref. MTM2011-23204, Ministerio de Ciencia e Innovación | Ref. PID2020-118101GB-I00, Xunta de Galicia | Ref. 10PXIB 300 068 PR
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Regression analysis with spatially-varying coefficients using generalized additive models (GAMs)

  • Lopez Alvarez, Francisco De Asis
  • Roca Pardiñas, Javier
  • Ordóñez Galán, Celestino
Regression models for spatial data have attracted the attention of researchers from different fields given their widespread application. In this work we analyze the utility of generalized additive models (GAMs) as regression methods with spatially-dependent coefficients. Particularly, three different aspects of the regression analysis were addressed: model definition and estimation, testing spatial heterogeneity, and variable selection. Spatial heterogeneity was addressed through bootstrapping, while and algorithm using the Bayesian Information Criterion (BIC) was implemented for variable selection to reduce computation time. In addition, this study makes a comparison of GAMs with two of the most common methods for regression with spatially-varying coefficients: Geographically Weighted Regression (GWR) and Multiscale Geographically Weighted Regression (MGWR), using both synthetic and real data., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Explainable generalized additive neural networks with independent neural network training

  • Ortega Fernández, Inés
  • Sestelo Perez, Marta
  • Martinez Villanueva, Nora
Neural Networks are one of the most popular methods nowadays given their high performance on diverse tasks, such as computer vision, anomaly detection, computer-aided disease detection and diagnosis or natural language processing. While neural networks are known for their high performance, they often suffer from the so-called “black-box” problem, which means that it is difficult to understand how the model makes decisions. We introduce a neural network topology based on Generalized Additive Models. By training an independent neural network to estimate the contribution of each feature to the output variable, we obtain a highly accurate and explainable deep learning model, providing a flexible framework for training Generalized Additive Neural Networks which does not impose any restriction on the neural network architecture. The proposed algorithm is evaluated through different simulation studies with synthetic datasets, as well as a real-world use case of Distributed Denial of Service cyberattack detection on an Industrial Control System. The results show that our proposal outperforms other GAM-based neural network implementations while providing higher interpretability, making it a promising approach for high-risk AI applications where transparency and accountability are crucial., Xunta de Galicia, Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Influencia de las estrategias de promoción en la captación y adherencia al remo federado, The influence of promotional strategies of recruitmaent and adherence to federated rowing

  • García González, Ivan
  • Iglesias Pérez, Maria Carmen
  • Vicente Vila, Pedro
El objetivo principal del presente estudio es conocer la eficacia de las estrategias de promoción del Club de Remo Vila de Cangas a la iniciación a la práctica del remo en la población escolar, analizando la eficacia de dichas estrategias en la captación a un evento deportivo escolar y la influencia de la captación sobre una posterior adherencia al deporte federado. Por otra parte, se investiga el efecto que las variables independientes género, categoría de edad, modalidad de participación y titularidad del centro escolar presentan en la captación y adherencia de practicantes del remo. La muestra analizada está conformada por las participaciones de 5.820 escolares de E.P. y E.S.O. procedentes de 13 centros educativos. Las tres estrategias de captación utilizadas se denominan Conoce mi club, Conoce el remo de banco fijo y Conoce el remo, todas incluidas dentro del programa Xogade de promoción del deporte escolar de la Xunta de Galicia e implementadas entre los años 2014 y 2019. El análisis descriptivo de los datos se realiza mediante tablas de contingencia, frecuencias y porcentajes; para el análisis bivariante se emplea el contraste de independencia Chi-Cuadrado de Pearson y la Regresión Logística Binaria al objeto de analizar la asociación entre las variables establecidas y estimar las odd ratios (OR) ajustadas. La utilización de estrategias de promoción obtiene un nivel de captación del 15,5% y una tasa de adherencia global del 1,8%. Dicha tasa de adherencia muestra un valor del 0,9% entre los no captados. Sin embargo, si los escolares son previamente captados la tasa de adherencia llega a valores del 6,4%. Se encontró una asociación estadísticamente significativa entre la captación y las actividades de promoción (p<0,05) y entre la adherencia y la captación (p<0,05), que se mantiene en cada uno de los géneros, categorías, modalidades de participación y tipos de centro. La regresión logística binaria manifiesta que la actividad de promoción B, las categorías benjamín y alevín, la modalidad no inclusiva y la titularidad privado-concertada están positiva y significativamente asociadas con ser captado. Respecto a la adherencia, el ser previamente captado, el género femenino y la titularidad pública se asocian positiva y significativamente con adherirse al remo federado., The main objective of this study is to know the effectiveness of the promotional strategies of the Vila de Cangas Rowing Club in the initiation to the practice of rowing in the school population. We analyze the effectiveness of these strategies in the recruitment of participants to a school sporting event, and the influence of recruitment on subsequent adherence to federated sport. On the other hand, we investigate the effect of the independent variables gender, age category, mode of participation and ownership of the school on the recruitment and adherence of rowing practitioners. The analyzed sample consisted of 5,820 primary and secondary school pupils from 13 schools. The three recruitment strategies used are called: “Know my club”, “Know the fixed-bench rowing” and “Know the rowing”, all included within the Xogade programme for the promotion of school sports of the Xunta de Galicia and implemented between 2014 and 2019. The descriptive analysis of the data is carried out using contingency tables, frecuencies and percentages. For the bivariate analysis, the Pearson Chi-Square test of independence and Binary Logistic Regression are used in order to study the association between the established variables and estimate the adjusted odds ratios (OR)., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Nonparametric estimation of the shape functions and related asymptotic results

  • Jiménez Gamero, María Dolores
  • Pardo Fernández, Juan Carlos
Arriaza et al (Metrika 82:99–124, 2019) introduced the right and left shape functions, which enjoy interesting properties in terms of describing the global form of a distribution. This paper proposes and studies nonparametric estimators of those functions. The estimators involve nonparametric estimation of the quantile and density functions. Pointwise and uniform consistency are proved under general regularity assumptions, as well as the limit in law. Simulations are included to study the practical performance of the proposed estimators. The analysis of a real data set illustrates the methodology., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Universidade de Vigo/CISUG
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Testing for an ignorable sampling bias under random double truncation

  • De Uña Alvarez, Jacobo
In clinical and epidemiological research doubly truncated data often appear. This is the case, for instance, when the data registry is formed by interval sampling. Double truncation generally induces a sampling bias on the target variable, so proper corrections of ordinary estimation and inference procedures must be used. Unfortunately, the nonparametric maximum likelihood estimator of a doubly truncated distribution has several drawbacks, like potential nonexistence and nonuniqueness issues, or large estimation variance. Interestingly, no correction for double truncation is needed when the sampling bias is ignorable, which may occur with interval sampling and other sampling designs. In such a case the ordinary empirical distribution function is a consistent and fully efficient estimator that generally brings remarkable variance improvements compared to the nonparametric maximum likelihood estimator. Thus, identification of such situations is critical for the simple and efficient estimation of the target distribution. In this article, we introduce for the first time formal testing procedures for the null hypothesis of ignorable sampling bias with doubly truncated data. The asymptotic properties of the proposed test statistic are investigated. A bootstrap algorithm to approximate the null distribution of the test in practice is introduced. The finite sample performance of the method is studied in simulated scenarios. Finally, applications to data on onset for childhood cancer and Parkinson’s disease are given. Variance improvements in estimation are discussed and illustrated., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Tesis doctoral (DoctoralThesis)

The k-Sample problem for truncated and censored data, El problema de las k Muestras para datos truncados y censurados, O problema das k Mostras para datos truncados e censurados

  • Lago Balseiro, Adrian
The two-sample Kolmogorov-Smirnov (KS) statistic is a classical method to compare two populations. In Survival Analysis, the classical KS method is not applicable in general because the samples may suffer from random truncation or random censoring. The main goal of this thesis is to adapt the classic KS test to left truncation. Properties of the proposed test will be studied, theoretically and through simulations. A bootstrap resampling plan will be proposed to approximate the null distribution of the statistic in practice and the goodness of such approximation will be studied as well. Possible extensions of the method will be studied, such as the extension to a k-sample scenario and the possibility to include right censorship or double truncation. A Cramér-von Mises and Anderson-Darling type statistics will be studied as well., El estadístico de Kolmogorov-Smirnov (KS) para el problema de las dos muestras es un método clásico para comparar dos poblaciones. En Análisis de Supervivencia, el test de KS clásico no es aplicable en general porque las muestras podrían sufrir de truncamiento o censura aleatorios. El objetivo principal de esta tesis doctoral es adaptar el test clásico de KS a truncamiento por la izquierda. Se estudiarán las propiedades del test propuesto, tanto teóricas como en escenarios simulados. Se propondrá un plan de remuestreo bootstrap para aproximar la distribución nula del estadístico de contraste en la práctica, y se estudiará la bondad de tal aproximación. Se estudiarán, como posibles extensiones del método, el escenario con k muestras y la posibilidad de incorporar censura por la derecha o truncamiento doble. También se estudiarán otros posibles estadísticos, como los tipo Cramér-von Mises y Anderson-Darling., O estatístico de Kolmogorov-Smirnov (KS) é un método clásico para comparar dúas poboacións. En Análise de Supervivencia, o método de KS clásico non é aplicable en xeral, porque as mostras poden sufrir de truncamento ou censura aleatorios. O principal obxectivo desta tese é adaptar o test de KS clásico a truncamento pola esquerda. As propiedades do test proposto serán estudadas, tanto teoricamente coma mendiante simulacións. Un plan de remostraxe bootstrap será proposto para aproximar a distribución do estatístico baixo a hipótese nula na práctica e tamén se estudará a bondade de dita aproximación. Estudaranse tamén posibles extensións do método, como a extensión a escenarios con k mostras e a posibilidade de incluír censura pola dereita ou truncamento dobre. Estudaranse tamén estatísticos tipo Cramér-von Mises e Anderson-Darling., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Agencia Estatal de Investigación | Ref. PID2023-148811NB-I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Cox regression with doubly truncated responses and time-dependent covariates: the impact of innovation on firm survival

  • De Uña Alvarez, Jacobo
  • Martínez Senra, Ana Isabel
  • Otero Giraldez, Maria Soledad
  • Quintás Corredoira, María de los Ángeles
The creation of new firms is an important incentive for the economic growth of a country, since it generates employment, it encourages the competition, and promotes innovation. In this work, we investigate the survival of Spanish firms which were created since 2001 and closed down between 2004 and 2012. The information was gathered from Technological Innovation Panel (PITEC), a survey with a focus the technological innovation in Spanish firms. In particular, a Cox regression model with time-dependent covariates was used in order to identify and quantify the determinants of the risk of exit for the firm. The selection bias due to the interval sampling for the firms was corrected by using methods for doubly truncated lifetimes. Interestingly, it is seen how the correction for the selection bias changes both the size and the statistical significance of the effects provided by standard Cox regression., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Agencia Estatal de Investigación | Ref. PID2019-106677GB-I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

A generalized additive model (GAM) approach to principal component analysis of geographic data

  • Lopez Alvarez, Francisco De Asis
  • Ordóñez Galán, Celestino
  • Roca Pardiñas, Javier
Geographically Weighted Principal Component Analysis (GWPCA) is an extension of classical PCA to deal with the spatial heterogeneity of geographical data. This heterogeneity results in a variance–covariance matrix that is not stationary but changes with the geographical location. Despite its usefulness, this method presents some unsolved issues, such as finding an appropriate bandwidth (size of the vicinity) as a function of the retained components. In this work, we address the problem of calculating principal components for geographical data from a new perspective that overcomes this problem. Specifically we propose a scale-location model which uses generalized additive models (GAMs) to calculate means for each variable and a correlation matrix that relates the variables, both depending on the spatial location. It should be noticed that although we deal with geographic data, our methodology cannot be considered strictly spatial since we assume that there is not a spatial correlation structure in the error term. Our approach does not require to calculate an optimal bandwidth as a function of the number of components retained in the analysis. Instead, the covariance matrix is estimated using smooth functions adapted to the data, so the smoothness can be different for each element of the matrix. The proposed methodology was tested with simulated data and compared with GWPCA. The result was a better representation of the data structure in the proposed method. Finally, we show the possibilities of our method in a problem with real data regarding air pollution and socioeconomic factors., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Universidad de Oviedo | Ref. 2017∕00001∕006∕001∕097
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Deciphering the plot preferences of forest contractors when purchasing stumpage through conjoint analysis

  • Alonso Martinez, Laura
  • Picos Martín, Juan
  • Iglesias Pérez, Maria Carmen
  • Iglesias González, David
  • Armesto González, Julia
In order to have sustainable wood value chains, it is essential to understand the factors that determine the incorporation of wood resources into them. Forest contractors are among the key players in this process. This study evaluates the preferences of forest contractors when purchasing forest parcels. The variables considered are: slope of the terrain, distance to the nearest road, parcel size, parcel shape, and the fragmentation of the surrounding land. The study area is Galicia, a region in northwestern Spain that is dominated by small-scale family forestry. An Attribute Levels Survey was designed to establish threshold values of the considered variables to afterwards perform a Choice Based Conjoint Analysis (CBCA). The CBCA allowed to analyze the preferences of forest contractors in relation to the values of these variables and their relative importance. Also, it allowed to generate a map of the level of preference for all the Galician forest parcels with individual, private ownership. The most noteworthy result of this survey was that size greatly impacts the preference of timber contractors, preceded by slope and followed by distance to roads. These results will aid in the design of landscape-scale policies in a geospatial dimension, like the promotion of forest associations, and will lead to an improvement in the sustainability of wood supply., Agencia Estatal de Investigación | Ref. PID2019-111581RB-I00, Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Agencia Estatal de Investigación | Ref. FPU19/02054, Universidade de Vigo/CISUG
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Nonparametric approaches for direct approximation of the spatial quantiles

  • García Soidán, Maria Del Pilar Hortensia
  • Cotos Yáñez, Tomas Raimundo
The estimation of the spatial quantiles provides information on the thresholds of a spatial variable. This methodology is particularly appealing for its application to data of pollutants, so as to assess their level of risk. A spatial quantile can be approximated through different mechanisms, proposed in the statistics literature, although these approaches suffer from several drawbacks, regarding their lack of optimality or the fact of not leading to direct approximations. Thus, the current work introduces alternative procedures, which try to overcome the aforementioned issues by employing order statistics, similarly as done for independent data. With this aim, the available observations are appropriately transformed to yield a sample of the process at each target site, so that the data obtained are then ordered and used to derive the spatial quantile at the corresponding location. The new methodology can be directly applied to data from processes that are either stationary or that deviate from this condition for a non-constant trend and, additionally, it can be even extended to heteroscedastic data. Simulation studies under different scenarios have been accomplished, whose results show the adequate performance of the proposed estimators. A further step of this research is the application of the new approaches to data of nitrogen dioxide concentrations, to exemplify the potential of the quantile estimates to check the thresholds of a pollutant at a specific moment, as well as their evolution over time., Agencia Estatal de Investigación | Ref. PID2020-113979RB-C22, Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Xunta de Galicia | Ref. ED431C 2019/25, Universidade de Vigo/CISUG
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

A method for determining groups in cumulative incidence curves in competing risk data

  • Sestelo Perez, Marta
  • Meira Machado, Luis
  • Martinez Villanueva, Nora
  • Roca Pardiñas, Javier
The cumulative incidence function is the standard method for estimating the marginal probability of a given event in the presence of competing risks. One basic but important goal in the analysis of competing risk data is the comparison of these curves, for which limited literature exists. We proposed a new procedure that lets us not only test the equality of these curves but also group them if they are not equal. The proposed method allows determining the composition of the groups as well as an automatic selection of their number. Simulation studies show the good numerical behavior of the proposed methods for finite sample size. The applicability of the proposed method is illustrated using real data., Fundação para a Ciência e a Tecnologia | Ref. UIDB/00013/2020, Fundação para a Ciência e a Tecnologia | Ref. UIDP/00013/2020, Agencia Estatal de Investigación | Ref. PID2020‐118101GB‐I00
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Goodness‐of‐fit testing for a regression model with a doubly truncated response

  • De Uña Alvarez, Jacobo
In survival analysis and epidemiology, among other fields, interval sampling is often employed. With interval sampling, the individuals undergoing the event of interest within a calendar time interval are recruited. This results in doubly truncated event times. Double truncation, which may appear with other sampling designs too, induces a selection bias, so ordinary statistical methods are generally inconsistent. In this paper, we introduce goodness-of-fit procedures for a regression model when the response variable is doubly truncated. With this purpose, a marked empirical process based on weighted residuals is constructed and its weak convergence is established. Kolmogorov–Smirnov– and Cramér–von Mises–type tests are consequently derived from such core process, and a bootstrap approximation for their practical implementation is given. The performance of the proposed tests is investigated through simulations. An application to model selection for AIDS incubation time as depending on age at infection is provided., Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Universidade de Vigo/CISUG
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Tesis doctoral (DoctoralThesis)

Machine learning approaches and explainability for real-time cyberattack detection, Aprendizaxe automática e explicabilidade para a detección en tempo real de ciberataques, Aprendizaje automático y explicabilidad para la detección de ciberataques en tiempo real

  • Ortega Fernández, Inés
The emergence of Industry 4.0, also known as connected industry, has blurred the boundary between IT (Information Technology) and OT (Operational Technology) domains. This integration has led to increased exposure of industrial environments to cyberattacks, data privacy risks, and software system vulnerabilities, significantly expanding the number of potential entry points for cyberattackers. Each connected device, be it a sensor, actuator, or control system, can become a vulnerability if not appropriately secured or monitored. Besides, the complexity and diversity of the systems in Industry 4.0 environments make it difficult to detect and respond to security incidents. Traditional security solutions may not be effective against sophisticated attacks designed to exploit industrial systems; in particular, Denial-of-service (DoS) attacks pose one of the most significant threats to Industrial Control Systems (ICS), as they can impact both the communication network and the cyber-physical layer. A common approach to securing information networks is Intrusion Detection Systems (IDS). Anomalybased IDS have gained popularity in recent years due to their ability to detect and mitigate unknown or zero-day attacks, whereas signature-based IDS can only respond to known attacks. Industrial systems operate under vastly different conditions compared to traditional networks, with much stricter requirements for real-time performance and availability. However, they usually exhibit a stable pattern of behaviour that can be leveraged by machine learning algorithms for the detection of advanced cyberattacks. This thesis, which is presented as a compendium of articles, has as its primary goal the research and development of novel machine learning methods for detecting advanced cyberattacks and intrusions in industrial networks, with a focus on their unique conditions and characteristics. The research begins by examining various types of Denial of Service (DoS) attacks and their effects on smart grids, a type of industrial control system where IT integration with the electrical distribution network enhances power distribution and management. Through this investigation, the impact of advanced DoS attacks in the industry was assessed, setting them as a promising research path given the ease of implementation of the attack, their applicability to different industrial settings, and the high impact on the behaviour of legitimate devices. Subsequently, an advanced Intrusion Detection System (IDS) was designed, evaluated and validated, enabling near real-time processing and analysis of industrial network traffic. This IDS has three principal features: its capability for near real-time, passive network traffic collection that does not interfere with industrial devices’ operations; the employment of unsupervised neural network techniques for anomaly detection, outperforming other state-of-the-art methods; and the validation of our approach in a real industrial setting within the Galician food industry. This validation assessed the IDS’s false positive rate and the practicality of the proposed IDS architecture in a near real-time context. Furthermore, this thesis introduces a novel approach to provide complete explainability of the decision-making processes of deep neural networks, enabling the determination of how each variable influences the detection of cyber attacks. This approach utilises a neural network architecture based on Generalized Additive Models, resulting in a deep learning model that is both highly accurate and interpretable, surpassing other GAM-based neural network implementations in terms of interpretability while having a similar accuracy. This method has been released as an open-source package for the software environment for statistical computing R to promote reproducible research and facilitate continuous development and improvement of the developed methodology., O crecemento da Industria 4.0, ou industria conectada, ten diluído a fronteira entre o ámbito IT e OT: as redes operativas atópanse cada vez máis expostas a ciberataques, riscos contra a privacidade dos datos e vulnerabilidades nos sistemas software. Un enfoque común á seguridade das redes de información son os Sistemas de Deteccion de Intrusións (IDS, polas súas siglas en inglés). Os IDS baseados en anomalías popularizáronse nos últimos anos grazas á súa habilidade para detectar e mitigar ataques descoñecidos ou de día 0, mentres que os baseados en firmas tan só poden dar resposta a ataques coñecidos. Os sistemas industriais traballan baixo condicións moi diferentes das redes tradicionais: os requerimentos de tempo real e dispoñibilidade son moito máis estritos, pero presentan un patrón de comportamento estable que pode ser aproveitado por algoritmos de aprendizaxe automática. Esta tese de doutoramento ten como principal obxectivo desenvolver novos métodos basados en aprendizaxe automático para a detección de ciberataques e intrusións en redes industriais, que teñan en consideración as condicións e características particulares destas redes. Este estudo propón o deseño, análise e evaluación dun IDS indutrial que permita o procesado de tráfico de rede industrial en case tempo real. O IDS caracterizarase por ofrecer tres funcionalidades clave: a capacidade de recoller tráfico de rede en case tempo real de forma pasiva, sen interrumpir a operación dos dispositivos industriais; o emprego de novos métodos baseados no uso de redes neuronáis (tanto supervisados como non supervisados) para determinar se o comportamento actual é anómalo, e o estudo de métodos que permitan explicar os motivos que levaron á detección do ataque a partir das variables que influíron na decisión dos algoritmos. Os enfoques actuáis á detección de anomalías e explicabilidade en redes neuronáis necesitan ser analizadas para establecer unha liña base de estudo, asegurando a novidade da proposta. Os métodos propostos deben ser capaces de determinar, a partires de datos estadísticos capturados de forma pasiva na rede, se o estado actual é normal ou anómalo. Os resultados deben ser explicables e entendibles por un operador humano, facilitando a interpretación dos resultados e a toma de decisións. A decisión sobre as técnicas de aprendizaxe automático a empregar realizarase partindo dunha análise en profundidade das aproximacións actuáis á detección de anomalías e métodos estadísticos clásicos, seleccionando aquelas máis axeitadas para os entornos industriais. As diferencias entre os ámbitos industriais e de información deben ser descritas formalmente para decidir o mellor conxunto de características a empregar. Por último, os patróns de ataque máis comúns no ámbito industrial deben ser identificados para determinar as variables que deberá capturar e analizar o sistema. Ademáis, cremos firmemente na difusión dos métodos propostos e na súa posible aplicación en outras áreas de coñecemento, polo que se distribuirán os métodos desenvolvidos en forma de paquetes R. Por último, a metodoloxía proposta aplicarase nun escenario real de ciberseguridade industrial para a súa validación., El crecimiento de la Industria 4.0, o industria conectada, ha difuminado la frontera entre el ámbito de TI y OT: las redes operativas están cada vez más expuestas a ciberataques, riesgos contra la privacidad de los datos y vulnerabilidades en los sistemas de software. Un enfoque común para la seguridad de las redes de información son los Sistemas de Detección de Intrusiones (IDS, por sus siglas en inglés). Los IDS basados en anomalías se han popularizado en los últimos años debido a su capacidad para detectar y mitigar ataques desconocidos o de día cero, mientras que los basados en firmas solo pueden responder a ataques conocidos. Los sistemas industriales trabajan bajo condiciones muy diferentes a las redes tradicionales: los requisitos de tiempo real y disponibilidad son mucho más estrictos, pero presentan un patrón de comportamiento estable que puede ser aprovechado por algoritmos de aprendizaje automático. Esta tesis de doctorado tiene como objetivo principal desarrollar nuevos métodos basados en aprendizaje automático para la detección de ciberataques e intrusiones en redes industriales, teniendo en cuenta las condiciones y características particulares de estas redes. Este estudio propone el diseño, análisis y evaluación de un IDS industrial que permita el procesamiento de tráfico de red industrial casi en tiempo real. El IDS se caracterizará por ofrecer tres funcionalidades clave: la capacidad de recolectar tráfico de red casi en tiempo real de forma pasiva, sin interrumpir la operación de los dispositivos industriales; el empleo de nuevos métodos basados en el uso de redes neuronales (tanto supervisados como no supervisados) para determinar si el comportamiento actual es anómalo, y el estudio de métodos que permitan explicar los motivos que llevaron a la detección del ataque a partir de las variables que influyeron en la decisión de los algoritmos. Los enfoques actuales de detección de anomalías y explicabilidad en redes neuronales deben ser analizados para establecer una línea base de estudio, asegurando la novedad de la propuesta. Los métodos propuestos deben ser capaces de determinar, a partir de datos estadísticos capturados de forma pasiva en la red, si el estado actual es normal o anómalo. Los resultados deben ser explicables y comprensibles para un operador humano, facilitando la interpretación de los resultados y la toma de decisiones. La decisión sobre las técnicas de aprendizaje automático a emplear se realizará a partir de un análisis en profundidad de las aproximaciones actuales a la detección de anomalías y métodos estadísticos clásicos, seleccionando aquellos más adecuados para los entornos industriales. Las diferencias entre los ámbitos industriales y de información deben ser descritas formalmente para decidir el mejor conjunto de características a emplear. Por último, los patrones de ataque más comunes en el ámbito industrial deben ser identificados para determinar las variables que el sistema deberá capturar y analizar. Además, creemos firmemente en la difusión de los métodos propuestos y su posible aplicación en otras áreas de conocimiento, por lo que los métodos desarrollados se distribuirán en forma de paquetes R. Por último, la metodología propuesta se aplicará en un escenario real de ciberseguridad industrial para su validación., Axencia Galega de Innovación | Ref. IN854A 2019/15, Agencia Estatal de Investigación | Ref. PID2020-118101GB-I00, Xunta de Galicia
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

TwoSampleTest.HD: An R Package for the Two-Sample Problem with High-Dimensional Data

  • Cousido-Rocha, Marta
  • de Uña-Álvarez, Jacobo
The two-sample problem refers to the comparison of two probability distributions via two independent samples. With high-dimensional data, such comparison is performed along a large number p of possibly correlated variables or outcomes. In genomics, for instance, the variables may represent gene expression levels for p locations, recorded for two (usually small) groups of individuals. In this paper we introduce TwoSampleTest.HD, a new R package to test for the equal distribution of the p outcomes. Specifically, TwoSampleTest.HD implements the tests recently proposed by (Cousido- Rocha, Uña-Álvarez, and Hart 2019) for the low sample size, large dimensional setting. These tests take the possible dependence among the p variables into account, and work for sample sizes as small as two. The tests are based on the distance between the empirical characteristic functions of the two samples, when averaged along the p locations. Different options to estimate the variance of the test statistic under dependence are allowed. The package TwoSampleTest.HD provides the user with individual permutation p-values too, so feature discovery is possible when the null hypothesis of equal distribution is rejected. We illustrate the usage of the package through the analysis of simulated and real data, where results provided by alternative approaches are considered for comparison purposes. In particular, benefits of the implemented tests relative to ordinary multiple comparison procedures are highlighted. Practical recommendations are given., The authors acknowledge financial support from the Grant PID2020-118101GB-I00, Ministerio de Ciencia e Innovación., Peer reviewed