Statistically motivated quantifiers GUHA matrices/searches can be seen from a statistical point of view, too. We may ask ‘Is the coincidence of two predicates.

Slides:



Advertisements
Similar presentations
Introductory Mathematics & Statistics for Business
Advertisements

Tests of Hypotheses Based on a Single Sample
Lesson Tests about a Population Parameter.
Statistical Inferences Based on Two Samples
Testing Hypotheses About Proportions
January Structure of the book Section 1 (Ch 1 – 10) Basic concepts and techniques Section 2 (Ch 11 – 15): Inference for quantitative outcomes Section.
Hypothesis Testing. To define a statistical Test we 1.Choose a statistic (called the test statistic) 2.Divide the range of possible values for the test.
Key Stone Problems… Key Stone Problems… next Set 11 © 2007 Herbert I. Gross.
1 Chapter 9 Hypothesis Testing Developing Null and Alternative Hypotheses Type I and Type II Errors One-Tailed Tests About a Population Mean: Large-Sample.
1 1 Slide STATISTICS FOR BUSINESS AND ECONOMICS Seventh Edition AndersonSweeneyWilliams Slides Prepared by John Loucks © 1999 ITP/South-Western College.
Chapter 10 Section 2 Hypothesis Tests for a Population Mean
Testing Hypotheses About Proportions Chapter 20. Hypotheses Hypotheses are working models that we adopt temporarily. Our starting hypothesis is called.
Introduction to Hypothesis Testing
QM Spring 2002 Business Statistics Introduction to Inference: Hypothesis Testing.
Lecture 2: Thu, Jan 16 Hypothesis Testing – Introduction (Ch 11)
Introduction to Hypothesis Testing
Visual Recognition Tutorial
Inference about a Mean Part II
Experimental Evaluation
APPLICATIONS OF DIFFERENTIATION 4. In Sections 2.2 and 2.4, we investigated infinite limits and vertical asymptotes.  There, we let x approach a number.
McGraw-Hill/IrwinCopyright © 2009 by The McGraw-Hill Companies, Inc. All Rights Reserved. Chapter 9 Hypothesis Testing.
Overview Definition Hypothesis
Copyright © 2005 Brooks/Cole, a division of Thomson Learning, Inc Chapter 9 Introduction to Hypothesis Testing.
Hypothesis Testing.
Copyright © 2013, 2010 and 2007 Pearson Education, Inc. Chapter Inference on the Least-Squares Regression Model and Multiple Regression 14.
Copyright © Cengage Learning. All rights reserved. 10 Inferences Involving Two Populations.
1 CSI5388: Functional Elements of Statistics for Machine Learning Part I.
LIMITS AND DERIVATIVES 2. In Sections 2.2 and 2.4, we investigated infinite limits and vertical asymptotes.  There, we let x approach a number.  The.
Random Regressors and Moment Based Estimation Prepared by Vera Tabakova, East Carolina University.
Maximum Likelihood Estimator of Proportion Let {s 1,s 2,…,s n } be a set of independent outcomes from a Bernoulli experiment with unknown probability.
10.2 Tests of Significance Use confidence intervals when the goal is to estimate the population parameter If the goal is to.
Lecture 16 Section 8.1 Objectives: Testing Statistical Hypotheses − Stating hypotheses statements − Type I and II errors − Conducting a hypothesis test.
1 Chapter 8 Hypothesis Testing 8.2 Basics of Hypothesis Testing 8.3 Testing about a Proportion p 8.4 Testing about a Mean µ (σ known) 8.5 Testing about.
McGraw-Hill/Irwin Copyright © 2007 by The McGraw-Hill Companies, Inc. All rights reserved. Chapter 8 Hypothesis Testing.
Ex St 801 Statistical Methods Inference about a Single Population Mean.
Inen 460 Lecture 2. Estimation (ch. 6,7) and Hypothesis Testing (ch.8) Two Important Aspects of Statistical Inference Point Estimation – Estimate an unknown.
Understanding Basic Statistics Fourth Edition By Brase and Brase Prepared by: Lynn Smith Gloucester County College Chapter Nine Hypothesis Testing.
We will now study some special kinds of non-standard quantifiers. Definition 4. Let  (x),  (x) be two fixed formulae of a language L n such that x is.
Hypothesis Testing Steps : 1. Review Data : –Sample size. –Type of data. –Measurement of data. –The parameter ( ,  2,P) you want to test. 2. Assumption.
Copyright (c) 2004 Brooks/Cole, a division of Thomson Learning, Inc. Chapter 7 Inferences Concerning Means.
Chapter 9 Hypothesis Testing Understanding Basic Statistics Fifth Edition By Brase and Brase Prepared by Jon Booze.
Chapter Nine Hypothesis Testing.
Virtual University of Pakistan
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Business Statistics Topic 7
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Significance Tests: A Four-Step Process
Hypothesis Testing: Hypotheses
Chapter 6 Hypothesis tests.
Chapter 9 Hypothesis Testing.
Chapter 9 Hypothesis Testing.
LESSON 20: HYPOTHESIS TESTING
Comparing two Rates Farrokh Alemi Ph.D.
Virtual University of Pakistan
Copyright © Cengage Learning. All rights reserved.
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Last Update 12th May 2011 SESSION 41 & 42 Hypothesis Testing.
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Chapter 9: Testing a Claim
Unit 5: Hypothesis Testing
Chapter 9: Testing a Claim
Significance Tests: The Basics
Statistical inference for the slope and intercept in SLR
Presentation transcript:

Statistically motivated quantifiers GUHA matrices/searches can be seen from a statistical point of view, too. We may ask ‘Is the coincidence of two predicates  (x) and  (x) just random or is there some statistically justified dependence between them’? For example, it is custom to use the  2 test to compare observed and expected values; a genetic experiment might hypothesise that the next generation of plants will exhibit a certain set of colours. By comparing the observed results with the expected ones, we can decide whether our original hypothesis is valid. We will study two statistically motivated quantifiers in details and mention several others Fisher quantifier  , 0 <   0.5. Fisher quantifier corresponds to the test of hypothesis Probability(  (x)|  (x)) > Probability(  (x)|  (x)) with significance . For example, our data may concern health and smoking. Let v(  (x)) = TRUE mean ‘x is a smoker’ and v(  (x))) = TRUE mean ‘x has cancer. If an output of a GUHA procedure is  (x)  0.05  (x), we accept the hypothesis ‘Smoking causes cancer’ and doing so there is a 0.05 probability that we make a mistake. More precisely, a Fisher quantifier   (on the level , 0 <   0.5) is defined such that, for any model M, v((  (x)    (x) ) = TRUE iff ad > bc and I

Theorem 9. Fisher quantifier is associational. Proof. Consider models M 0, M 1, M 2, M 3, M 4 such that and (i) v M 0 ((  (x)    (x) ) = TRUE. We should show that (ii) v M 1 ((  (x)    (x) ) = TRUE, (iii) v M 2 ((  (x)    (x) ) = TRUE, (iv) v M 3 ((  (x)    (x) ) = TRUE, (v) v M 4 ((  (x)    (x) ) = TRUE. However, since Fisher quantifiers are invariant under interchanging b and c and under interchanging a and d, we have to prove (ii) and (iii) only. Assume (i). Then First we notice bc  ad +d, which holds true by assumption.

Second, we notice that, for each i = 0,…,min{b,c} Therefore Trivially (a+1)d > bc. Therefore (ii) holds. Next consider the model M 2 and the value 1° Let b  c. First notice that which holds true by assumption. Second, notice that, for all i = 0,…,b-1 we have Obviously, the last inequality is true. We may now estimate B:

2° Let c < b [i.e. c  b-1]. Again and, for each i = 0,…,c, Trivially, ad > (b-1)c. This completes the proof. Theorem 10.             and     are sound rules of inference for Fisher quantifiers Proof. The claim becomes obvious as soon as we realise that, for any model M, interchanging (  and  ) or (  and ,  and  ) has no effect on the values

Lehman proved in 1959 that Fisher test is the most powerful in the class of unbiased tests of the null hypothesis   0 and the alternative hypothesis  > 0. On the other hand, the computation of the Fisher test for larger m is complicated, the complexity of computation increasing rapidly. For this practical reason, another test, the  2 test is widely used. This test is only asymptotical, but the approximation is rather good for reasonable cardinalities (a, b, c, d  5, m d  20). We will see that Fisher quantifier and  2 quantifier have similar properties. For the exact definition of the  2 quantifiers, recall the following: let a continuous one- dimensional distribution function D(x) be given. For each  [0,1], the value D -1 (  ) is called the  -quantile of D. The  2 quantifier (on the level  ) is defined such that v(  (x)    (x)) = TRUE iff ad > bc and where is the (1-  ) quantile of the  2 -distribution function. In practice, an  2 -association rule  (x)    (x)) corresponds to a test (on the level  ) of the null hypothesis of independence of  (x) and  (x) against the alternative one of the positive dependence. Theorem 10.  2 -quantifiers are associational. Proof. The  2 -quantifiers are invariant under interchanging b and c and under inter- changing a and d. Thus, it is enough to show that if (i) v M 0 ((  (x)    (x) ) = TRUE, then (ii) v M 1 ((  (x)    (x) ) = TRUE (iii) v M 2 ((  (x)    (x) ) = TRUE, too, where I

First we realise that, for any numbers A, B, x, y greater than 0, it holds that Thus, in particular Since b 2 c 2 = bcbc  abcd, we have b 2 c 2 (r + k + 1)  abcd(r + k + 1). Thus, to prove (*) it is sufficient to prove Substituting r = a + b, k = a + c results (by Maple)

We have now proved the inequality Therefore we have We conclude Trivially (a+1)d > bc holds. We summarise: if v M 0 ((  (x)    (x) ) = TRUE, then v M 1 ((  (x)    (x) ) = TRUE. Next consider an inequality (*) is equivalent to the following inequality: The right hand side of (**) is obviously  1. Moreover, since the left hand side of (**) is  1. Therefore (**) holds and, hence, (*) holds, too. Trivially ad > (b-1)c. We have shown: if v M 0 ((  (x)    (x) ) = TRUE, then v M 2 ((  (x)    (x) ) = TRUE, too. This completes the proof. Theorem 11. For  2 -quantifiers             and     are sound rules of inference. Proof. The claim is obvious as, for any model M, interchanging (  and  ) or (  and ,  and  ) has no effect on the values

Exercises. Some more statistically motivated quantifiers. Show that the following quantifiers are implicational. 23. Lower critical implication  ! p,  Base, where 0 0. v(  (x)  ! p,  Base  (x)) = TRUE iff An association rule  (x)  ! p,  Base  (x) corresponds to a test (on the level of  ) of a null hypothesis H 0 : P(Suc|Ant)  p against the alternative one H 1 : P(Suc|Ant) > p. If the association rule  (x)  ! p,  Base  (x) is true in data matrix M then the alternative hypothesis is accepted. 24. Upper critical implication  ? p,  Base, where 0 0. v(  (x)  ? p,  Base  (x)) = TRUE iff An association rule  (x)  ? p,  Base  (x) corresponds to a test (on the level of  ) of a null hypothesis H 0 : P(Suc|Ant)  p against the alternative one H 1 : P(Suc|Ant) < p. If the association rule  (x)  ? p,  Base  (x) is true in data matrix M then the alternative hypothesis is accepted.

Show that the following quantifiers are associational. 24. Double lower critical implication  ! p,  Base, 0 0. v(  (x)  ! p,  Base  (x)) = TRUE iff 25. Double upper critical implication  ? p,  Base, 0 0. v(  (x)  ? p,  Base  (x)) = TRUE iff

Show that the following quantifiers are associational. 26. Lower critical equivalence  ! p,  Base, 0 0. v(  (x)  ! p,  Base  (x)) = TRUE iff 27. Upper critical implication  ? p,  Base, 0 0. v(  (x)  ? p,  Base  (x)) = TRUE iff