DOCSLIB.ORG

Big Data Effective Processing and Analysis of Very Large and Unstructured Data for Official Statistics

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Big Data Effective Processing and Analysis of Very Large and Unstructured Data for Official Statistics Big Data Effective Processing and Analysis of Very Large and Unstructured data for Official Statistics. Big Data and Interference in Official Statistics Part 2 Diego Zardetto Istat ([email protected]) Eurostat Big Data: Methodological Pitfalls Outline • Representativeness (w.r.t. the desired target population) Selection Bias Actual target population unknown Often sample units’ identity unclear/fuzzy Pre processing errors (acting like measurement errors in surveys) Social media & sentiment analysis: pointless babble & social bots • Ubiquitous correlations Causation fallacy Spurious correlations • Structural break in Nowcasting Algorithms Eurostat 1 Social Networks: representativeness (1/3) The “Pew Research Center” carries out periodic sample surveys on the way US people use the Internet According to its Social Media Update 2013 research: • Among US Internet users aged 18+ (as of Sept 2013) 73% of online adults use some Social Networks 71% of online adults use Facebook 18% of online adults use Twitter 17% of online adults use Instagram 21% of online adults use Pinterest 22% of online adults use LinkedIn • Subpopulations using these Social platforms turn out to be quite different in terms of: Gender, Ethnicity, Age, Education, Income, Urbanity Eurostat Social Networks: representativeness (2/3) According to Social Media Update 2013 research • Social platform users turn out to belong to typical social & demographic profiles Pinterest is skewed towards females LinkedIn over-represents higher education and higher income Twitter and Instagram are biased towards younger adults, urban dwellers, and non-whites Facebook too is skewed towards younger people (though much less than Twitter and Instagram) and females • More interestingly, even looking at all such subpopulations as a whole, Social Networks users turn out to constitute a biased sample of US online adults • Big Data generated via Social Networks are very likely to be non representative of the US general population Eurostat 2 Social Networks: representativeness (3/3) Incidence of Social Network users among US Internet adults within specific social/demographical groups – Sept 2013 • superscript letters (e.g. a) indicate statistically significant differences between cells Eurostat Tweets: the relevance issue There is growing evidence that a big portion of tweets lack any relevant information • And even more so in the Official Statistics perspective Such tweets are categorized as Pointless Babble • e.g. “I am eating a sandwich now” Pear analytics “Twitter Study 2009” estimated that about 40% of tweets were Pointless Babble Recent extensive studies conducted by Statistics Netherlands on millions of Dutch Tweets revealed that more than 50% had no statistical interest Challenge : how to filter out such a big noise in order to uncover meaningful signals inside huge amounts of Tweets? Eurostat 3 How many Twitter users are actually humans? According to her Twitter bio, Carina Santos ( @scarina91 ) is a young, blonde haired, girl based in Rio de Janeiro, working as a journalist at Globo, a 24-hour news channel in Brazil • She sends around 50 tweets per day on a variety of popular topics (spanning from sport news to gossip) to her nearly 1000 followers • Twitalyzer, a social media analytics firm, judged her to be more influential than online personalities like Oprah Winfrey or the Dalai Lama Eurostat How many Twitter users are actually humans? Eurostat 4 How many Twitter users are actually humans? According to her Twitter bio, Carina Santos ( @scarina91 ) is a young, blonde haired, girl based in Rio de Janeiro, working as a journalist at Globo, a 24-hour news channel in Brazil • She sends around 50 tweets per day on a variety of popular topics (spanning from sport news to gossip) to her nearly 1000 followers • Twitalyzer, a social media analytics firm, judged her to be more influential than online personalities like Oprah Winfrey or the Dalai Lama Carina Santos is not a human but rather a social bot created in 2011 by a team of Brazilian computer science researchers: • a software robot able to interact and communicate with humans by following appropriate social behaviors and rules “She” constantly checked Twitter API restrictions to avoid being identified as a bot and blocked by Twitter Maximum number of requests per hour is 350 Each user can follow up to 2,000 users A user cannot follow more than 1,000 users per day Eurostat Tweets: the Social Bots issue Security analysts and artificial intelligence researchers agree that the current amount of Twitter bots is big and growing fast • It has been observed that thousands of dormant Twitter profiles suddenly began posting hundreds of messages during last years social turmoil and political crises around the world (e.g. Syrian civil war and Arab Spring protests in Egypt, Libya, Tunisia) PeekAnalytics estimated in 2011 that only 35% of Twitter users’ followers were real people Maybe this estimate is not accurate or even grossly exaggerated, but the problem is still there: • How much of what can be mined from Tweets can safely be used to analyze, interpret or predict people behavior? Challenge : beyond removing the pointless bubble noise, how to discriminate bot generated Tweets deliberately intended to mimic meaningful human signals? How to adjust analyses for bot traffic? Eurostat 5 Inference in the Official Statistics Realm Outline Traditional paradigm • Top-down: data are planned • Traditional inference approaches Design based survey sampling theory Model-assisted approach Model based inference The need of a new paradigm to deal with Big Data • Bottom-up: data are already there • Exploratory analysis / Knowledge discovery approach Algorithmic inference: data mining techniques, machine learning, … Eurostat Design Based Inference (1/2) Probability sampling • Samples are drawn by means of rigorously random algorithms • Every unit in the population has a known, non-zero probability of being selected in the random sample Data sampling is entirely controlled Ideally, THE STATISTICIAN IS THE ONE AND THE ONLY RANDOMIZER Ideally means ignoring all non-sampling errors (e.g. list problems, total and item nonresponse, measurement errors…) This approach allows to: • Build unbiased estimators (or nearly so) even if samples are not naively representative, because we can adjust for unequal inclusion probabilities • Exploit probability theory to assess the quality of obtained estimates Eurostat 6 Design Based Inference (2/2) U = {1,2,..., N} s = {1,2,...,n} s ∈ S p: S → ]1,0[ p() s = Pr( s is selected ) π k = Pr( k ⊂ s) = ∑ p() s s⊃ k π kj = Pr({ k,} j ⊂ s) = ∑ p() s s⊃{,} k j ˆ YHT = ∑dk y k = ∑ yk /1( π k ) k∈ s k∈ s ∆ π π ˆ ˆ kj k j V (YHT ) = ∑∑dk y k dj y j = ∑∑1− dk y k dj y j k∈ s j ∈ s π kj k∈ s j ∈ s π kj Eurostat Model Assisted Inference (1/2) Reference framework is still Design Based • Observed values (y, x) of interest variable Y and auxiliary variables x are deterministic , non-random quantities • Relations (if any) between Y and X are generated by Nature (i.e. by real- world, domain-specific phenomena) Information about the target population is available from sources external to the survey at hand • Can use this information to describe relations between Y and X through a model Model Assisted Inference is a suite of methods to improve the quality of Design Based inferences by hinging upon available auxiliary information in a systematic and rigorous way build more efficient (but still nearly unbiased) estimators reduce nonresponse bias The model is assisting only (i.e. descriptive): no stochastic structure ever assumed! Eurostat 7 Model Assisted Inference (2/2) x ξ : yk ~ k ⋅β + ε k Eξ ()ε k = 0 ∀k ∈U 2 Vξ ()ε k = σ < ∞ ∀k ∈U Cov ξ (,)ε kε j = 0 ∀k, j ∈U ˆ x ˆ x ˆ ˆ X Xˆ ˆ YGREG = (∑ k )⋅β + ∑dk(y k − k ⋅β) = YHT + ( − HT )⋅β = ∑ wk y k k∈ U k∈ s k∈ s −1 X = x d d X Xˆ T−1 xt T−1 xt x ∑ k wk () = gk ()dk gk =1 + ( − HT )⋅ ⋅ k = ∑dk k ⋅ k k∈ U k∈ s ˆ ˆ ˆ ˆ ˆ e = y − x ⋅βˆ VY( GREG ) ≈ VY( GREG ,lin ) = V ∑dk(g k e k ) k k k k∈ s Eurostat Model Based Inference (1/2) Sampling design is non informative and can be ignored Equivalent to assuming that the data have been selected by simple random sampling with replacement Observed values (y, x) are regarded as realizations of random variables : Y (the response variable) and X (the explanatory variables) Assumptions are made on the probability distribution of Y and (sometimes) of X A statistical model linking the response variable Y to the explanatory variables X is proposed and tested against the observed data (y, x) • Often models describe how the conditional expectation of Y given X (or some nice/smart function of it, think of GLM) depends on X NATURE IS THE RANDOMIZER • A Model is a (human interpretable) guess on “how” Nature is working Eurostat 8 Model Based Inference (2/2) ξ : Y ~ N(X⋅β, σ 21) YX X Eξ ( ) = ⋅β YX 21 2 Vξ ( ) = σ σ < ∞ yˆ()x =x ⋅βˆ βˆ = (Xt X)−1 ⋅(Xt Y) ˆ x ˆ Yξ = ∑ yk + ∑ yˆk = ∑ yk + ∑ k ⋅β k∈ s k∈ ( U − s ) k∈ s k∈ ( U − s ) Eurostat Algorithmic Inference (1/2) Models are replaced by Algorithms , whose ultimate goal is to map X values (input) to Y values (predicted output) • Machine Learning : emphasis on prediction, i.e. learning data properties from a training set and being able to mimic “what” would happen to another data set • Data Mining : emphasis on data exploration and discovery (of correlations, rules, clusters, patterns, hidden structures, …) Fitting a Model is replaced by tuning
Load more

Recommended publications
  • Bayesian Inference for Two-Parameter Gamma Distribution Assuming Different Noninformative Priors
    Revista Colombiana de Estadística Diciembre 2013, volumen 36, no. 2, pp. 319 a 336 Bayesian Inference for Two-Parameter Gamma Distribution Assuming Different Noninformative Priors Inferencia Bayesiana para la distribución Gamma de dos parámetros asumiendo diferentes a prioris no informativas Fernando Antonio Moala1;a, Pedro Luiz Ramos1;b, Jorge Alberto Achcar2;c 1Departamento de Estadística, Facultad de Ciencia y Tecnología, Universidade Estadual Paulista, Presidente Prudente, Brasil 2Departamento de Medicina Social, Facultad de Medicina de Ribeirão Preto, Universidade de São Paulo, Ribeirão Preto, Brasil Abstract In this paper distinct prior distributions are derived in a Bayesian in- ference of the two-parameters Gamma distribution. Noniformative priors, such as Jeffreys, reference, MDIP, Tibshirani and an innovative prior based on the copula approach are investigated. We show that the maximal data information prior provides in an improper posterior density and that the different choices of the parameter of interest lead to different reference pri- ors in this case. Based on the simulated data sets, the Bayesian estimates and credible intervals for the unknown parameters are computed and the performance of the prior distributions are evaluated. The Bayesian analysis is conducted using the Markov Chain Monte Carlo (MCMC) methods to generate samples from the posterior distributions under the above priors. Key words: Gamma distribution, noninformative prior, copula, conjugate, Jeffreys prior, reference, MDIP, orthogonal, MCMC. Resumen En este artículo diferentes distribuciones a priori son derivadas en una in- ferencia Bayesiana de la distribución Gamma de dos parámetros. A prioris no informativas tales como las de Jeffrey, de referencia, MDIP, Tibshirani y una priori innovativa basada en la alternativa por cópulas son investigadas.
    [Show full text]
  • Bayes Estimation and Prediction of the Two-Parameter Gamma Distribution
    Bayes Estimation and Prediction of the Two-Parameter Gamma Distribution Biswabrata Pradhan ¤& Debasis Kundu y Abstract In this article the Bayes estimates of two-parameter gamma distribution is considered. It is well known that the Bayes estimators of the two-parameter gamma distribution do not have compact form. In this paper, it is assumed that the scale parameter has a gamma prior and the shape parameter has any log-concave prior, and they are independently distributed. Under the above priors, we use Gibbs sampling technique to generate samples from the posterior density function. Based on the generated samples, we can compute the Bayes estimates of the unknown parameters and also can construct highest posterior density credible intervals. We also compute the approximate Bayes estimates using Lindley's approximation under the assumption of gamma priors of the shape parameter. Monte Carlo simulations are performed to compare the performances of the Bayes estimators with the classical estimators. One data analysis is performed for illustrative purposes. We further discuss about the Bayesian prediction of future observation based on the observed sample and it is observed that the Gibbs sampling technique can be used quite e®ectively, for estimating the posterior predictive density and also for constructing predictive interval of the order statistics from the future sample. Keywords: Maximum likelihood estimators; Conjugate priors; Lindley's approximation; Gibbs sampling; Predictive density, Predictive distribution. Subject Classifications: 62F15; 65C05 Corresponding Address: Debasis Kundu, Phone: 91-512-2597141; Fax: 91-512-2597500; e-mail: [email protected] ¤SQC & OR Unit, Indian Statistical Institute, 203 B.T. Road, Kolkata, Pin 700108, India yDepartment of Mathematics and Statistics, Indian Institute of Technology Kanpur, Pin 208016, India 1 1 Introduction The two-parameter gamma distribution has been used quite extensively in reliability and survival analysis particularly when the data are not censored.
    [Show full text]
  • A Philosophical Framework for Explainable Artificial
    Reasons, Values, Stakeholders: A Philosophical Framework for Explainable Artificial Intelligence Atoosa Kasirzadeh [email protected] University of Toronto (Toronto, Canada) & Australian National University (Canberra, Australia) Abstract The societal and ethical implications of the use of opaque artificial intelligence systems for consequential decisions, such as welfare allocation and criminal justice, have generated a lively debate among multiple stakeholder groups, in- cluding computer scientists, ethicists, social scientists, policy makers, and end users. However, the lack of a common language or a multi-dimensional frame- work to appropriately bridge the technical, epistemic, and normative aspects of this debate prevents the discussion from being as productive as it could be. Drawing on the philosophical literature on the nature and value of explana- tions, this paper offers a multi-faceted framework that brings more conceptual precision to the present debate by (1) identifying the types of explanations that are most pertinent to artificial intelligence predictions, (2) recognizing the rel- evance and importance of social and ethical values for the evaluation of these explanations, and (3) demonstrating the importance of these explanations for incorporating a diversified approach to improving the design of truthful al- arXiv:2103.00752v1 [cs.CY] 1 Mar 2021 gorithmic ecosystems. The proposed philosophical framework thus lays the groundwork for establishing a pertinent connection between the technical and ethical aspects of artificial intelligence systems. Keywords: Ethical Algorithm, Explainable AI, Philosophy of Explanation, Ethics of Artificial Intelligence Preprint submitted to Elsevier 1. Preliminaries Governments and private actors are seeking to leverage recent develop- ments in artificial intelligence (AI) and machine learning (ML) algorithms for use in high-stakes decision making.
    [Show full text]
  • An Algorithmic Inference Approach to Learn Copulas Arxiv:1910.02678V1
    An Algorithmic Inference Approach to Learn Copulas Bruno Apolloni Department of Computer Science University of Milano Via Celoria 18, 20133, Milano, Italy [email protected] October 8, 2019 Abstract We introduce a new method for estimating the parameter α of the bivariate Clayton copulas within the framework of Algorithmic Inference. The method consists of a variant of the standard boot- strapping procedure for inferring random parameters, which we ex- arXiv:1910.02678v1 [stat.ML] 7 Oct 2019 pressly devise to bypass the two pitfalls of this specific instance: the non independence of the Kendall statistics, customarily at the basis of this inference task, and the absence of a sufficient statistic w.r.t. α. The variant is rooted on a numerical procedure in order to find the α estimate at a fixed point of an iterative routine. Although paired 1 with the customary complexity of the program which computes them, numerical results show an outperforming accuracy of the estimates. Keywords | Copulas' Inference, Clayton copulas, Algorithmic Infer- ence, Bootstrap Methods. 1 Introduction Copulas are the atoms of the stochastic dependency between variables. For short, the Sklar Theorem (Sklar 1973) states that we may split the joint cumulative distribution function (CDF) FX1;:::;Xn of variables X1;:::;Xn into the composition of their marginal CDF FXi ; i = 1; : : : ; n through their copula 1 CX1;:::;Xn . Namely : FX1;:::;Xn (x1; : : : ; xn) = CX1;:::;Xn (FX1 (x1);:::;FXn (xn)) (1) While the random variable FXi (Xi) is a uniform variable U in [0; 1] for whatever continuous FXi thanks to the Probability Integral Transform Theo- rem (Rohatgi 1976) (with obvious extension to the discrete case), CX1;:::;Xn (FX1 (X1); :::;FXn (Xn)), hence CX1;:::;Xn (U1;:::;Un); has a specific distribution law which characterizes the dependence between the variables.
    [Show full text]
  • Study On: Tools for Causal Inference from Cross-Sectional Innovation Surveys with Continuous Or Discrete Variables
    Study on: Tools for causal inference from cross-sectional innovation surveys with continuous or discrete variables Dominik Janzing June 28, 2016 Abstract The design of effective public policy, in innovation support, and in other areas, has been constrained by the difficulties involved in under- standing the effects of interventions. The limited ability of researchers to unpick causal relationships means that our understanding of the likely ef- fects of public policy remains relatively poor. If improved methods could be found to unpick causal connections our understanding of important phenomena and the impact of policy interventions could be significantly improved, with important implications for future social prosperity of the EU. Recently, a series of new techniques have been developed in the natu- ral and computing sciences that are enabling researchers to unpick causal links that were largely impossible to explore using traditional methods. Such techniques are currently at the cutting edge of research, and remain technical for the non-specialist, but are slowly diffusing to the broader academic community. Economic analysis of innovation data, in particu- lar, stands to gain a lot from applying techniques developed by the ma- chine learning community. Writing recently in the Journal of Economic Perspectives, Varian (2014, p3) explains \my standard advice to graduate students these days is go to the computer science department and take a class in machine learning. There have been very fruitful collaborations between computer scientists and statisticians in the last decade or so, and I expect collaborations between computer scientists and econometricians will also be productive in the future." However, machine learning tech- niques have not yet been applied to innvovation policy.
    [Show full text]
  • Predictive Inference for Non Probability Samples
    Discussion Paper Predictive inference for non-probability samples: a simulation study The views expressed in this paper are those of the author(s) and do not necessarily reflect the policies of Statistics Netherlands 2015 | 13 Bart Buelens Joep Burger Jan van den Brakel 3 CBS | Discussion Paper, 2015 | 13 Non-probability samples provide a challenging source of information for official statistics, because the data generating mechanism is unknown. Making inference from such samples therefore requires a novel approach compared with the classic approach of survey sampling. Design-based inference is a powerful technique for random samples obtained via a known survey design, but cannot legitimately be applied to non-probability samples such as big data and voluntary opt-in panels. We propose a framework for such non-probability samples based on predictive inference. Three classes of methods are discussed. Pseudo-design-based methods are the simplest and apply traditional design-based estimation despite the absence of a survey design; model-based methods specify an explicit model and use that for prediction; algorithmic methods from the field of machine learning produce predictions in a non-linear fashion through computational techniques. We conduct a simulation study with a real-world data set containing annual mileages driven by cars for which a number of auxiliary characteristics are known. A number of data generating mechanisms are simulated, and—in absence of a survey design—a range of methods for inference are applied and compared to the known population values. The first main conclusion from the simulation study is that unbiased inference from a selective non- probability sample is possible, but access to the variables explaining the selection mechanism underlying the data generating process is crucial.
    [Show full text]
  • Engines of Order: a Mechanology of Algorithmic Techniques
    A Mechanology Engines of Algorithmic of Order Techniques BERNHARD RIEDER Amsterdam University Press Engines of Order The book series RECURSIONS: THEORIES OF MEDIA, MATERIALITY, AND CULTURAL TECHNIQUES provides a platform for cuttingedge research in the field of media culture studies with a particular focus on the cultural impact of media technology and the materialities of communication. The series aims to be an internationally significant and exciting opening into emerging ideas in media theory ranging from media materialism and hardware-oriented studies to ecology, the post-human, the study of cultural techniques, and recent contributions to media archaeology. The series revolves around key themes: – The material underpinning of media theory – New advances in media archaeology and media philosophy – Studies in cultural techniques These themes resonate with some of the most interesting debates in international media studies, where non-representational thought, the technicity of knowledge formations and new materialities expressed through biological and technological developments are changing the vocabularies of cultural theory. The series is also interested in the mediatic conditions of such theoretical ideas and developing them as media theory. Editorial Board – Jussi Parikka (University of Southampton) – Anna Tuschling (Ruhr-Universität Bochum) – Geoffrey Winthrop-Young (University of British Columbia) Engines of Order A Mechanology of Algorithmic Techniques Bernhard Rieder Amsterdam University Press This publication is funded by the Dutch Research Council (NWO). Chapter 1 contains passages from Rieder, B. (2016). Big Data and the Paradox of Diversity. Digital Culture & Society 2(2), 1-16 and Rieder, B. (2017). Beyond Surveillance: How Do Markets and Algorithms ‘Think’? Le Foucaldien 3(1), n.p.
    [Show full text]
  • 9<HTMIPI=Aadajf>
    58 Engineering Springer News 6/2008 springer.com/booksellers S. Akella, N. Amato, W. Huang, B. Mishra (Eds.) B. Apolloni, University of Milano, Italy; W. Pedrycz, V. Botti, A. Giret, Technical University of Valencia, University of Alberta, Edmonton, Canada; S. Bassis, Spain Algorithmic Foundation of D. Malchiodi, University of Milano, Italy Robotics VII ANEMONA The Puzzle of Granular A Multi-agent Methodology for Holonic Selected Contributions of the Seventh Computing International Workshop on the Algorithmic Manufacturing Systems Foundations of Robotics “Rem tene, verba sequentur“ Gaius J. Victor, Rome ANEMONA is a multi-agent system (MAS) meth- VI century b.c. odology for holonic manufacturing system (HMS) Algorithms are a fundamental component of The ultimate goal of this book is to bring the analysis and design, based on HMS requirements. robotic systems: they control or reason about fundamental issues of information granularity, ANEMONA defines a mixed top-down and bottom- motion and perception in the physical world. inference tools and problem solving procedures up development process, and provides HMS-specific They receive input from noisy sensors, consider into a coherent, unified, and fully operational guidelines to help the designer in identifying and geometric and physical constraints, and operate on framework. The objective is to offer the reader a implementing holons. In ANEMONA, the specified the world through imprecise actuators. The design comprehensive, self-contained, and uniform expo- HMS is divided into concrete aspects that form and analysis of robot algorithms therefore raises a sure to the subject.The strategy is to isolate some different “views” of the system. unique combination of questions in control theory, fundamental bricks of Computational Intelligence The development process of ANEMONA provides computational and differential geometry, and in terms of key problems and methods, and discuss clear and HMS-specific modelling guidelines computer science.
    [Show full text]
  • An Algorithmic Information Calculus for Causal Discovery and Reprogramming Systems
    An Algorithmic Information Calculus for Causal Discovery and Reprogramming Systems Hector Zenil^*a,b,c,d,e, Narsis A. Kiani*a,b,d,e, Francesco Marabitab,d, Yue Dengb, Szabolcs Eliasb,d, Angelika Schmidtb,d, Gordon Ballb,d, & Jesper Tegnér^b,d,f a) Algorithmic Dynamics Lab, Center for Molecular Medicine, Karolinska Institutet, Stockholm, 171 76, Sweden b) Unit of Computational Medicine, Center for Molecular Medicine, Department of Medicine, Solna, Karolinska Institutet, Stockholm, 171 76, Sweden c) Department of Computer Science, University of Oxford, Oxford, OX1 3QD, UK. d) Science for Life Laboratory, Solna, 171 65, Sweden e) Algorithmic Nature Group, LABORES for the Natural and Digital Sciences, Paris, 75006, France. f) Biological and Environmental Sciences and Engineering Division, Computer, Electrical and Mathematical Sciences and Engineering Division, King Abdullah University of Science and Technology (KAUST), Thuwal 23955–6900, Kingdom of Saudi Arabia * Shared-first authors ^ Corresponding authors Author Contributions: HZ, NK and JT are responsible for the general design and conception. HZ and NK are responsible for data acquisition. HZ, NK, YD, FM, SE and AS contributed to data analysis. HZ developed the methodology, with key contributions from NK and JT. HZ undertook most of the numerical experiments, with YD and FM contributing. HZ, AS, GB and SE contributed the literature-based Th17 enrichment analysis. HZ and JT wrote the article, with key contributions from NK. Correspondence should be addressed to HZ: [email protected]
    [Show full text]
  • Large Scale Multinomial Inferences and Its Applications in Genome Wide Association Studies
    Large Scale Multinomial Inferences and Its Applications in Genome Wide Association Studies Chuanhai Liu and Jun Xie Abstract Statistical analysis of multinomial counts with a large number K of cat- egories and a small number n of sample size is challenging to both frequentist and Bayesian methods and requires thinking about statistical inference at a very funda- mental level. Following the framework of Dempster-Shafer theory of belief func- tions, a probabilistic inferential model is proposed for this “large K and small n” problem. The inferential model produces a probability triplet (p,q,r) for an asser- tion conditional on observed data. The probabilities p and q are for and against the truth of the assertion, whereas r = 1 p q is the remaining probability called the probability of “don’t know”. The new− inference− method is applied in a genome-wide association study with very high dimensional count data, to identify association be- tween genetic variants to the disease Rheumatoid Arthritis. 1 Introduction We consider statistical inference for the comparison of two large-scale multino- mial distributions. This problem is motivated by genome-wide association studies (GWAS) with very high dimensional count data, i.e., single nucleotide polymor- phism (SNP) data. SNPs are major genetic variants that may associate with common diseases such as cancer and heart disease. A SNP has three possible genotypes, wild type homozygous, heterozygous, and mutation (rare) homozygous. In genome-wide association studies, genotypes of over 500,000 SNPs are typically measured for dis- ease (case) and normal (control) subjects, resulting in a large amount of count data.
    [Show full text]
  • Tight Bounds for SVM Classification Error
    Tight bounds for SVM classification error B. Apolloni, S. Bassis, S. Gaito, D. Malchiodi Dipartimento di Scienze dell'Informazione, Universita degli Studi di Milano Via Comelico 39/41, 20135 Milano, Italy E-mail: {apolloni, bassis, gaito, malchiodi}@dsi.unimi.it Abstract- We find very tight bounds on the accuracy of a Sup- lies in finding a separating hyperplane, i.e. an h E H such port Vector Machine classification error within the Algorithmic that all the points with a given label belong to one of the two Inference framework. The framework is specially suitable for this half-spaces determined by h. kind of classifier since (i) we know the number of support vectors really employed, as an ancillary output of the learning procedure, In order to obtain such a h, an SVM computes first the and (ii) we can appreciate confidence intervals of misclassifying solution {°ia, . , a* } of a dual constrained optimization probability exactly in function of the cardinality of these vectors. problem As a result we obtain confidence intervals that are up to an order m m narrower than those in the a 1 supplied literature, having slight max Z different meaning due to the different approach they come from, °Zai-2 aiicmjyiYjxi-x (1) but the same operational function. We numerically check the t=1 i,i-=i covering of these intervals. m aiyi = 0 (2) I. INTRODUCTION i-l Support Vector Machines (SVM for short) [1] represent ajt > O i = 1 . .. ,m, (3) an operational tool widely used by the Machine Learning where - denotes the standard dot product in RI, and then community.
    [Show full text]
  • Algorithmic Inference of Two-Parameter Gamma Distribution Bruno Apolloni, Simone Bassis
    Algorithmic Inference of Two-Parameter Gamma Distribution Bruno Apolloni, Simone Bassis To cite this version: Bruno Apolloni, Simone Bassis. Algorithmic Inference of Two-Parameter Gamma Distribution. Com- munications in Statistics - Simulation and Computation, Taylor & Francis, 2009, 38 (09), pp.1950-1968. 10.1080/03610910903171821. hal-00520670 HAL Id: hal-00520670 https://hal.archives-ouvertes.fr/hal-00520670 Submitted on 24 Sep 2010 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Communications in Statistics - Simulation and Computation For Peer Review Only Algorithmic Inference of Two-Parameter Gamma Distribution Journal: Communications in Statistics - Simulation and Computation Manuscript ID: LSSP-2009-0057.R1 Manuscript Type: Original Paper Date Submitted by the 01-Jul-2009 Author: Complete List of Authors: Apolloni, Bruno; University of Milano, Department of Computer Science Bassis, Simone; University of Milano, Department of Computer Science Two-parameters Gamma distribution, Algorithmic inference, Keywords: Population bootstrap, Gibbs sampling, Adaptive rejection sampling, Maximum likelihood estimator We provide an estimation procedure of the two-parameter Gamma distribution based on the Algorithmic Inference approach. As a key feature of this approach, we compute the joint probability distribution of these parameters without assuming any prior.
    [Show full text]