Non-parametric (Distribution-free) approaches p188 CN

Similar documents
SEVERAL μs AND MEDIANS: MORE ISSUES. Business Statistics

Non-parametric tests, part A:

Nonparametric Statistics. Leah Wright, Tyler Ross, Taylor Brown

CHAPTER 17 CHI-SQUARE AND OTHER NONPARAMETRIC TESTS FROM: PAGANO, R. R. (2007)

PSY 307 Statistics for the Behavioral Sciences. Chapter 20 Tests for Ranked Data, Choosing Statistical Tests

Nonparametric statistic methods. Waraphon Phimpraphai DVM, PhD Department of Veterinary Public Health

Nonparametric Statistics

4/6/16. Non-parametric Test. Overview. Stephen Opiyo. Distinguish Parametric and Nonparametric Test Procedures

Parametric versus Nonparametric Statistics-when to use them and which is more powerful? Dr Mahmoud Alhussami

= 1 i. normal approximation to χ 2 df > df

CHI SQUARE ANALYSIS 8/18/2011 HYPOTHESIS TESTS SO FAR PARAMETRIC VS. NON-PARAMETRIC

Nonparametric Statistics Notes

DETAILED CONTENTS PART I INTRODUCTION AND DESCRIPTIVE STATISTICS. 1. Introduction to Statistics

Chapter 7 Comparison of two independent samples

3. Nonparametric methods

Rank-Based Methods. Lukas Meier

HYPOTHESIS TESTING II TESTS ON MEANS. Sorana D. Bolboacă

Data are sometimes not compatible with the assumptions of parametric statistical tests (i.e. t-test, regression, ANOVA)

Chapter 15: Nonparametric Statistics Section 15.1: An Overview of Nonparametric Statistics

Introduction and Descriptive Statistics p. 1 Introduction to Statistics p. 3 Statistics, Science, and Observations p. 5 Populations and Samples p.

Lecture Slides. Section 13-1 Overview. Elementary Statistics Tenth Edition. Chapter 13 Nonparametric Statistics. by Mario F.

Lecture Slides. Elementary Statistics. by Mario F. Triola. and the Triola Statistics Series

Unit 14: Nonparametric Statistical Methods

BIO 682 Nonparametric Statistics Spring 2010

Module 9: Nonparametric Statistics Statistics (OA3102)

Fish SR P Diff Sgn rank Fish SR P Diff Sng rank

Textbook Examples of. SPSS Procedure

Non-parametric Tests

STAT 135 Lab 9 Multiple Testing, One-Way ANOVA and Kruskal-Wallis

Nonparametric Methods

Nonparametric Location Tests: k-sample

16. Nonparametric Methods. Analysis of ordinal data

What is a Hypothesis?

Basic Business Statistics, 10/e

Chapter 18 Resampling and Nonparametric Approaches To Data

Introduction to Nonparametric Statistics

ST4241 Design and Analysis of Clinical Trials Lecture 9: N. Lecture 9: Non-parametric procedures for CRBD

Non-Parametric Statistics: When Normal Isn t Good Enough"

N Utilization of Nursing Research in Advanced Practice, Summer 2008

Analysis of variance (ANOVA) Comparing the means of more than two groups

Statistical Inference Theory Lesson 46 Non-parametric Statistics

Biostatistics 270 Kruskal-Wallis Test 1. Kruskal-Wallis Test

Contents. Acknowledgments. xix

Lecture 26. December 19, Department of Biostatistics Johns Hopkins Bloomberg School of Public Health Johns Hopkins University.

ST4241 Design and Analysis of Clinical Trials Lecture 7: N. Lecture 7: Non-parametric tests for PDG data

Agonistic Display in Betta splendens: Data Analysis I. Betta splendens Research: Parametric or Non-parametric Data?

1 ONE SAMPLE TEST FOR MEDIAN: THE SIGN TEST

Hypothesis Testing. Hypothesis: conjecture, proposition or statement based on published literature, data, or a theory that may or may not be true

Chapter 12. Analysis of variance

E509A: Principle of Biostatistics. (Week 11(2): Introduction to non-parametric. methods ) GY Zou.

Inferences About the Difference Between Two Means

Degrees of freedom df=1. Limitations OR in SPSS LIM: Knowing σ and µ is unlikely in large

Statistics Handbook. All statistical tables were computed by the author.

Data Analysis: Agonistic Display in Betta splendens I. Betta splendens Research: Parametric or Non-parametric Data?

The One-Way Repeated-Measures ANOVA. (For Within-Subjects Designs)

MATH Notebook 3 Spring 2018

Inferential Statistics

Chapter 8 Class Notes Comparison of Paired Samples

a. When a data set is not normally distributed, what should you try in order to appropriately make statistical tests on that data?

Statistics for Managers Using Microsoft Excel Chapter 10 ANOVA and Other C-Sample Tests With Numerical Data

STAT 135 Lab 8 Hypothesis Testing Review, Mann-Whitney Test by Normal Approximation, and Wilcoxon Signed Rank Test.

Lec 3: Model Adequacy Checking

My data doesn t look like that..

Types of Statistical Tests DR. MIKE MARRAPODI

NON-PARAMETRIC STATISTICS * (

One-way ANOVA Model Assumptions

Correlation and Regression

Workshop Research Methods and Statistical Analysis

I i=1 1 I(J 1) j=1 (Y ij Ȳi ) 2. j=1 (Y j Ȳ )2 ] = 2n( is the two-sample t-test statistic.

Non-parametric Hypothesis Testing

Statistics: revision

Lecture 10: Non- parametric Comparison of Loca6on. GENOME 560, Spring 2015 Doug Fowler, GS

GROUPED DATA E.G. FOR SAMPLE OF RAW DATA (E.G. 4, 12, 7, 5, MEAN G x / n STANDARD DEVIATION MEDIAN AND QUARTILES STANDARD DEVIATION

Nonparametric tests. Timothy Hanson. Department of Statistics, University of South Carolina. Stat 704: Data Analysis I

Tentative solutions TMA4255 Applied Statistics 16 May, 2015

Kruskal-Wallis and Friedman type tests for. nested effects in hierarchical designs 1

ANOVA - analysis of variance - used to compare the means of several populations.

Intro to Parametric & Nonparametric Statistics

Statistics for Managers Using Microsoft Excel Chapter 9 Two Sample Tests With Numerical Data

4/22/2010. Test 3 Review ANOVA

Lecture 7: Hypothesis Testing and ANOVA

Non-parametric methods

Statistiek I. Nonparametric Tests. John Nerbonne. CLCG, Rijksuniversiteit Groningen.

TABLES AND FORMULAS FOR MOORE Basic Practice of Statistics

Week 14 Comparing k(> 2) Populations

STATISTIKA INDUSTRI 2 TIN 4004

Lecture 14: ANOVA and the F-test

SBAOD Statistical Methods & their Applications - II. Unit : I - V

Glossary. The ISI glossary of statistical terms provides definitions in a number of different languages:

Introduction to Statistical Analysis

Exam details. Final Review Session. Things to Review

Formulas and Tables by Mario F. Triola

Design of Experiments. Factorial experiments require a lot of resources

Solutions exercises of Chapter 7

Data analysis and Geostatistics - lecture VII

Comparison of Two Population Means

Statistical Procedures for Testing Homogeneity of Water Quality Parameters

Selection should be based on the desired biological interpretation!

Transition Passage to Descriptive Statistics 28

Hypothesis testing, part 2. With some material from Howard Seltman, Blase Ur, Bilge Mutlu, Vibha Sazawal

Transcription:

Week 1: Introduction to some nonparametric and computer intensive (re-sampling) approaches: the sign test, Wilcoxon tests and multi-sample extensions, Spearman s rank correlation; the Bootstrap. (ch14 &15) [pp188-05] Non-parametric (Distribution-free) approaches p188 CN The standard classical approaches to statistical analysis we have discussed assume 1) data behave like SRS s from normal distributions ) constant variance when there are more than one population involved If there are violations of these assumptions we can try either of the following: 1) use a transformation to bring the data into line with our assumptions ) use a less restrictive procedure, which does not impose any particular parametric model upon the data (i.e. non-parametric or distribution-free methods) 1

We will discuss the following non-parametric procedures Design independent samples (CRD) Paired samples More than independent samples (CRD) More than blocked samples (RBD) Parametric Test Non-parametric Test Pooled t-test Wilcoxon s Rank Sum Test t-test on Sign test or differences Wilcoxon s (paired t-test) Signed Rank One-way ANOVA F-test Two-way ANOVA F-test Test Kruskal Wallace Test Friedman Test MINITAB commands: Stats > Nonparametrics

Wilcoxon Rank Sum test (also known as Mann-Whitney test) p188 CN Example p 188 CN We take a SRS of 6 students who are Botany majors, and another SRS of 5 students who are Psychology majors and find that their final course grades are: Botany majors Psychology majors C+ B B- C A- D- D+ B+ A+ C+ B+ Is there evidence of a difference in achievement levels between the two types of students? 3

Sol: Botany Rank Psychology Rank C+ 4.5 B 7 B- 6 C 3 A- 10 D- 1 D+ B+ 8.5 A+ 11 C+ 4.5 B+ 8.5 Rank Sum 4 = T1 4 = T H0: the two populations are identical H1: populations differ in location Test statistic T = sum of the ranks of the smaller of the two independent samples (or for either if n1 = n) In our example T = 4 From Wilcoxon s rank sum test table, p-value > 0.05 x = 0.10 and so no evidence of a difference at the 10% level. 4

MINITAB output Mann-Whitney Test and CI: Psy_ranks, Bot_ranks N Median Psy_ranks 5 4.500 Bot_ranks 6 7.50 Point estimate for ETA1-ETA is -.000 96.4 Percent CI for ETA1-ETA is (-7.500,.497) W = 4.0 Test of ETA1 = ETA vs ETA1 not = ETA is significant at 0.3153 The test is significant at 0.3131 (adjusted for ties) Note n i ( n 1 + n + 1) µ T = and i nn 1 ( n 1 + n + 1) σ = i = 1, and for Ti 1 large sample sizes the distribution of T1 is approximately normal and so the normal tables can be used to find approximate p-values. 5

Kruskal-Wallace Test p191 CN Example p191 Sample 1 3 Sol: 9 1 5 10 4 7 11 1 H0: The 3 distributions are identical H1: Not all the distributions are the same Sample 1 3 9(6) 1(9) () 5(4) 10(7) 4(3) 7(5) 11(8) 1(1) T i 15 4 6 6

Kruskal-Wallace Statistic 1 T H = i 3( N + 1) where T N( N + 1) n i s are i the rank sums, N = total sample size Reject if H > χ t 1, α In our example 1 15 4 6 H = + + 3(9 + 1) = 7. 9(9 + 1) 3 3 3 χ = 5.99 31,0.05 and so reject H0. MINITAB output Kruskal-Wallis Test: score versus group Kruskal-Wallis Test on score Ave group N Median Rank Z 1 3 7.000 5.0 0.00 3 11.000 8.0.3 3 3.000.0 -.3 Overall 9 5.0 H = 7.0 DF = P = 0.07 * NOTE * One or more small samples 7

Example p19 CN Paired data p19 CN Suppose we want to compare two newly marketed wines, RioSamba (RS) and CubaSalsa (CS). We have 10 tasters, each taster tastes each wine. They rate both wines on a 0-10 scale, with 10 being the best, yielding: Taster 1 3 4 5 6 7 8 9 10 RS 10 8 10 9 7 8 10 8 8 6 CS 8 8 7 5 6 8 4 5 3 7 Test the null hypothesis of equivalence of the wines. 8

Sign Test Taster 1 3 4 5 6 7 8 9 10 RS 10 8 10 9 7 8 10 8 8 6 CS 8 8 7 5 6 8 4 5 3 7 Diff + 0 +3 +4 +1 0 +6 +3 +5-1 Test statistic (X) is the number of positive differences (or negative differences). The smaller is a bit easier to work with. X~ Bin (8, 0.5) In this example, the number negative differences = 1 p value= P( X 1) = (0.0039 + 0.0313) = 0.07 - This test is quite wasteful of information as it throws away the magnitudes of the differences and only retains the signs. 9

Wilcoxon s signed rank test p 193 CN Taster 1 3 4 5 6 7 8 9 10 RS 10 8 10 9 7 8 10 8 8 6 CS 8 8 7 5 6 8 4 5 3 7 Diff + 0 +3 +4 +1 0 +6 +3 +5-1 Rank 3 4.5 6 1.5 8 4.5 7 1.5 Let T+ = sum of ranks corresponding to the differences that were positive = 34.5 T- = sum of ranks corresponding to the differences that were negative = 1.5 For non-directional tests (i.e. two-sided tests) use the test statistic T =smaller of T+ or T- and the tables for the Wilcoxon s signed rank test gives PT ( 3) 0.05 and so p-value 0.05. If the alternative is one-sided: H1: RS has higher ratings than CS then T- which is 3 has p-value less than 0.05/ = 0.05 and so we reject H0. 10

Note: if n 5, under H0, T+ (or T-) has approximately a normal distribution with µ = nn ( + 1) nn ( + 1)(n + 1) and σ =. 4 4 Eg in this example with n=8 (this is not greater than 5) and T- =1.5 Z = -.31 and from standard normal tables, p-value = x 0.01 = 0.0 11

Friedman Test for a RBD p194 CN H0: The distributions for the k treatments within each block are identical H1: at least two distributions differ in location Example p194 CN Treatment 1 3 Block 1 9 1 Block 5 10 4 Block 3 7 11 1 Sol Treatment 1 3 Block 1 9() 1(3) (1) Block 5() 10(3) 4(1) Block 3 7() 11(3) 1(1) Rank sum 6 9 3 1

Test statistic: 1 F = Tt 3 b( t + 1) bt( t + 1) Reject H0 if F > χ t 1, α In our example F = 1 3 3(3+ 1) = 6 (36 81 9) + + 3 3(3+ 1) 31,0.05 χ = 5.99 and so we reject H0. Note: There are tables exact critical values available which are recommended for small sample sizes (like this example) Results for: Eg_Friedman'Test.MTW Friedman Test: Response versus Treatment blocked by Block S = 6.00 DF = P = 0.050 Sum of Treatment N Est Median Ranks 1 3 7.000 6.0 3 11.000 9.0 3 3 1.000 3.0 Grand median = 6.333 13

r S = 1 Rank correlation p 195 6 d i nn ( 1) Two students were asked to rate 8 textbooks for a course on a scale 0-0. The data are shown below: Row Textbook Student1 Student 1 A 4 4 B 10 6 3 C 18 0 4 D 0 14 5 E 1 16 6 F 8 7 G 5 11 8 H 9 7 Calculate the rank correlation between the two ratings and test whether there is a linear correlation between the two ratings. 14

Sol: Data Display Row Textbook Student1 Student rk1 rk d d 1 A 4 4 1 1 1 B 10 6 5 3 9 3 C 18 0 7 8-1 1 4 D 0 14 8 6 4 5 E 1 16 6 7-1 1 6 F 8 1 4-3 9 7 G 5 11 3 5-4 8 H 9 7 4 3 1 1 Descriptive Statistics: d Variable Sum d 30.00 r S =0.643 Test H 0 : ρ S = 0 H 1 : ρ S 0 t = r S n 1 r S =.057~ t n = t with 6 df 15