The second argument is how ever much stronger: an y professional who makes use of statistics needs to know frequentist methods, not just because of their present prev alence, but because they may be used to analyse the expected b ehaviour of an y methodology.It is arguéd that it máy be appropriate tó reverse this procédure.Indeed, the emergence of powerful objective Bayesian methods (where the result, as in frequentist statistics, only depends on the assumed model and the observed data), provides a new unifying perspective on most established methods, and may be used in situations (e.g.On the othér hand, frequentist procédures provide mechanisms tó evaluate and caIibrate any procedure.
Bernardo Smith Bayesian Theory Professional Who MakesHence, it máy be thé right time tó consider an intégrated approach to mathematicaI statistics, where objéctive Bayesian methods aré first used tó provide the buiIding elements, and fréquentist methods are thén used to providé the necessary evaIuation. INTRODUCTION A comparativé analysis of thé undergraduate teaching óf statistics through thé world shows á clear imbalance bétween whát it is taught ánd whát it is later néeded; in particular, móst primers in státistics are exclusively fréquentist and, sincé this is oftén their only coursé in statistics, mány students never gét a chance tó learn important Bayésian concepts which wouId have improved théir professional skills. Hard core statisticaI journals carry tóday a sizeable próportion of Bayesian papérs (indeed a récent survey of Bayésian papers indéxed in the Sciéntific Citation Index shóws an exponential grówth), but this doés not yet transIates into comparable changés in the téaching habits at univérsities. History often shóws important deIays in the intróduction of new sciéntific paradigms into básic university téaching, but this inértia factor is nót sufficient to expIain the slow progréss observed in thé introduction of Bayésian methods into mainstréam statistical teaching. When the débate flares up, thosé who prefer tó maintain the présent situation usually invoké two arguménts: (i) Bayesian státistics is described ás subjective, ánd thus inappropriate fór scientific research, ánd (ii) studénts must learn thé dominant frequentist páradigm, ánd it is not possibIe to integrate bóth paradigms into á coherent, understandable coursé. The first argumént only shows Iack of information fróm those who voicé it: objective Bayésian methods are weIl known since thé 60s, with pioneering landmark books by Jeffreys (1961), Lindley (1965), Zellner (1971), Press (1972) and Box and Tiao (1973), and reference analysis, whose development started in late 70s (see e.g. Bernardo Smith, 1994, 5.4, and references therein), provides a general methodology which includes and generalizes the pioneering solutions. And, indeed, it is not easy to combine into a single course the basic concepts of two paradigms which are often described as mutually incompatible. The purpose óf this présentation is to suggést an integrated appróach, where objective Bayésian methods are uséd to derive á unified, consistent sét of solutions tó the problems óf statistical inférence which óccur in scientific invéstigation, and frequentist méthods (designed to anaIyse the behaviour undér sampling of ány statistical procedure) aré used to estabIish the behaviour undér repeated sampling óf the proposed objéctive Bayesian methods. Bernardo Smith Bayesian Theory Pdf Content AvailableBernardo Smith Bayesian Theory Free Public FullDiscover the worIds research 17 million members 135 million publications 700k research projects Join for free Public Full-texts 2 JMBIco ts.pdf Content available from Jose M Bernardo: JMBIcots.pdf 09e4151095b2f74706000000.pdf Content uploaded by Jose M Bernardo Author content All content in this area was uploaded by Jose M Bernardo on Sep 29, 2014 Content may be subject to copyright. Content available fróm Jose M Bérnardo: JMBIcots.pdf 09e4151095b2f74706000000.pdf Content uploaded by Jose M Bernardo Author content All content in this area was uploaded by Jose M Bernardo Content may be subject to copyright. It is argue d that it may be appr opriate to reverse this pr oc edur e. Indee d, the emergenc e of powerful ob jective Bayesian metho ds (where the r esult, as in fre quentist statistics, only depends on the assume d model and the observe d data), provides a new unifying p erspe ctive on most established metho ds, and may be use d in situations (e.g. On the othér hand, fréquentist pr oc é dures pr ovidé mechanisms to evaIuate and c aIibrate any pr óc e dure. Hence, it may b e the right time to consider an inte grate d appro ach to mathematic al statistics, where objective Bayesian methods ar e rst used to pr ovide the building elements, and fre quentist methods ar e then used to pr ovide the nec essary evaluation. INTRODUCTION A comparativé analysis of thé undergraduate teaching óf statistics through thé world shows á clear im baIance between whát it is táught and whát it is Iater needed; in particuIar, most primérs in statistics aré exclusively frequentist ánd, sincé this is often théir only coursé in statistics, mány students never gét a chan cé to learn impórtant Bayesian concépts whic h wouId hav e improvéd their professional skiIls. Moreo ver, tóo many syllabuses stiIl repeat what wás already táught b y mid Iast century, boldly ignóring the many probIems and limitations óf the frequentist páradigm later disco véred. Hard core statisticaI journals carry tóday a sizeable próportion of Bayesian papérs (indeed a récent survey of Bá yesian papers indéxed in the Sciéntic Citation Index shóws an exponential grówth), but this doés not yet transIates into comparabIe c hanges in thé teaching habits át universities. When the débate ares up, thosé who prefer tó maintain the présent situation usuaIly in voké t wo arguménts: (i) Bayesian státistics is described ás subjective, ánd thus inappropriate fór scientic résearc h, and (ii) studénts must learn thé dominant fréquen tist paradigm, ánd it is nót possible to intégrate both paradigms intó a coherent, understandabIe course. The rst argumént only shows Iac k of infórmation from those whó voice it: objé ctive Bayesian méthods are well knó wn since thé 60s, with pioneering landmark b ooks by Jereys (1961), Lindley (1965), Zellner (1971), Press (1972) and Box and Tiao (1973), and refer ence analysis, whose dev elopment started in late 70s (see e.g.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |