Results 1 
9 of
9
Statistical Themes and Lessons for Data Mining
, 1997
"... Data mining is on the interface of Computer Science and Statistics, utilizing advances in both disciplines to make progress in extracting information from large databases. It is an emerging field that has attracted much attention in a very short period of time. This article highlights some statist ..."
Abstract

Cited by 36 (3 self)
 Add to MetaCart
Data mining is on the interface of Computer Science and Statistics, utilizing advances in both disciplines to make progress in extracting information from large databases. It is an emerging field that has attracted much attention in a very short period of time. This article highlights some statistical themes and lessons that are directly relevant to data mining and attempts to identify opportunities where close cooperation between the statistical and computational communities might reasonably provide synergy for further progress in data analysis.
Ménage a ̀ Trois Inference Style: Unifying Three Hypothesis Testing Doctrines∗
"... Three prominent ‘schools ’ of hypothesis testing exist, propelled by Fisher, Jeffreys and Neyman. Fisher extolled the virtue of the pvalue, whose magnitude signals the strength of evidence in the null hypothesis, H0. In contrast, Jeffreys’ approach favours the use of objective posterior probabiliti ..."
Abstract
 Add to MetaCart
Three prominent ‘schools ’ of hypothesis testing exist, propelled by Fisher, Jeffreys and Neyman. Fisher extolled the virtue of the pvalue, whose magnitude signals the strength of evidence in the null hypothesis, H0. In contrast, Jeffreys’ approach favours the use of objective posterior probabilities using a Bayesian framework, whilst Neyman resorted to fixed error probabilities, namely the computation of Type I and Type II errors. Here a unified framework of the competing doctrines is offered, using a new conditioning statistic which accommodates the pvalue density under the alternative hypothesis for both simple and composite tests. Critical pvalue curves and surfaces are derived to quickly allow conclusions to be drawn.
Toward a Bayesian DecisionTheoretic Approach to HypothesisTesting in Psychology
"... The testing of null hypotheses is a methodologically limited way to making decisions in psychology and science more generally. Null hypothesis significance testing, or “NHST, ” has undergone severe criticism since its inception with R.A. Fisher in early 20th century. The mere fact that the probabili ..."
Abstract
 Add to MetaCart
(Show Context)
The testing of null hypotheses is a methodologically limited way to making decisions in psychology and science more generally. Null hypothesis significance testing, or “NHST, ” has undergone severe criticism since its inception with R.A. Fisher in early 20th century. The mere fact that the probability of some data is less than some conventional cutoff (e.g.,.05) should not by itself suggest the rejection of the null and the adoption of some hypothesized alternative. What is missing from the conventional hypothesistesting paradigm is the incorporation of actual or estimated losses and costs that result from the possibilities of making erroneous decisions when choosing between H0 and H1. In this article, the problems with NHST are briefly reviewed, followed by a presentation of the Bayesian decisiontheoretic approach to hypothesis testing popularized by such decision theorists as Berger (1993) and Winkler (2002). The approach incorporates estimates of the actual losses incurred in the decisionmaking process, and is recommended for use in psychology and related sciences. Simple illustrations of the approach are given.
Toward a Bayesian DecisionTheoretic 1 Running Head: BAYESIAN DECISIONTHEORETIC APPROACH Toward a Bayesian DecisionTheoretic Approach to HypothesisTesting in Psychology
"... Toward a Bayesian DecisionTheoretic 2 The testing of null hypotheses is a methodologically limited way to making decisions in psychology and science more generally. Null hypothesis significance testing, or “NHST,” has undergone severe criticism since its inception with R.A. Fisher in early 20 th ce ..."
Abstract
 Add to MetaCart
(Show Context)
Toward a Bayesian DecisionTheoretic 2 The testing of null hypotheses is a methodologically limited way to making decisions in psychology and science more generally. Null hypothesis significance testing, or “NHST,” has undergone severe criticism since its inception with R.A. Fisher in early 20 th century. The mere fact that the probability of some data is less than some conventional cutoff (e.g.,.05) should not by itself suggest the rejection of the null and the adoption of some hypothesized alternative. What is missing from the conventional hypothesistesting paradigm is the incorporation of actual or estimated losses and costs that result from the possibilities of making erroneous decisions when choosing between H0 and H1. In this article, the problems with NHST are briefly reviewed, followed by a presentation of the Bayesian decisiontheoretic approach to hypothesis testing popularized by such decision theorists as Berger (1993) and Winkler (2002). The approach incorporates estimates of the actual losses incurred in the decisionmaking process, and is recommended for use in psychology and related sciences. Simple illustrations of the approach are given.
The pvalue, the Bayes/NeymanPearson Compromise and the Teaching of Statistical Inference in Introductory Business Statistics
"... Traditionally the NeymanPearson approach to hypothesis testing has been presented in introductory business statistics courses. However, many students as well as researchers find the decisions reached by this approach, i.e., reject/failtoreject, inconsistent with their understanding of the scienti ..."
Abstract
 Add to MetaCart
Traditionally the NeymanPearson approach to hypothesis testing has been presented in introductory business statistics courses. However, many students as well as researchers find the decisions reached by this approach, i.e., reject/failtoreject, inconsistent with their understanding of the scientific process, namely accumulating evidence in support of a hypothesis. The proposed framework provides an easily understood rationale for introducing the student to I.J. Good's Bayes/NeymanPearson compromise as represented by Good's standardized pvalues. Standardized pvalues are a useful and practical tool for the evidentialist interpretation of data within the context of NeymanPearson hypothesis testing, something
c ○ 1997 Kluwer Academic Publishers. Manufactured in The Netherlands. Statistical Themes and Lessons for Data Mining
, 1996
"... Abstract. Data mining is on the interface of Computer Science and Statistics, utilizing advances in both disciplines to make progress in extracting information from large databases. It is an emerging field that has attracted much attention in a very short period of time. This article highlights some ..."
Abstract
 Add to MetaCart
Abstract. Data mining is on the interface of Computer Science and Statistics, utilizing advances in both disciplines to make progress in extracting information from large databases. It is an emerging field that has attracted much attention in a very short period of time. This article highlights some statistical themes and lessons that are directly relevant to data mining and attempts to identify opportunities where close cooperation between the statistical and computational communities might reasonably provide synergy for further progress in data analysis.
RESEARCH ARTICLE Open Access
"... Interpretation of evidence in data by untrained ..."
(Show Context)
The Statistical Education of Harold Jeffreys
, 2005
"... The paper considers the statistical work of the physicist Harold Jeffreys. In 1933–4 Jeffreys had a controversy with R.A. Fisher, the leading statistician of the time. Prior to the encounter, Jeffreys had worked on probability as the basis for scientific inference and had used methods from the theor ..."
Abstract
 Add to MetaCart
The paper considers the statistical work of the physicist Harold Jeffreys. In 1933–4 Jeffreys had a controversy with R.A. Fisher, the leading statistician of the time. Prior to the encounter, Jeffreys had worked on probability as the basis for scientific inference and had used methods from the theory of errors in astronomy and seismology. He had also started to rework the theory of errors on the basis of his theory of probability. After the encounter Jeffreys produced a fullscale Bayesian treatment of statistics in the form of his Theory of Probability.
RESEARCH ARTICLE Open Access
"... Interpretation of evidence in data by untrained ..."
(Show Context)