Chapter 12 Nonparametric Statistics Understandable Statistics Ninth Edition By Brase and Brase Prepared by Yixun Shi Bloomsburg University of Pennsylvania Nonparametric Situations • At times, we will not know anything about the distributions of the populations from which we are sampling. • Recall that all of our inference techniques thus far have assumed either a normal or binomial distribution from the populations of interest. Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 2 Nonparametric Tests • Advantages: – Easy to apply – Quite general in nature • Disadvantages: – Wastes information – Accept the null hypothesis more often than with other tests – Less sensitive Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 3 The Sign Test • We wish to compare two populations. – Populations are not independent Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 4 Sign Test Method Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 5 Sign Test Method Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 6 Sign Test Method Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 7 Rank-Sum Test • Data values from the two populations are not paired. • Normal assumptions are not satisfied, or are at least questionable. Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 8 Rank-Sum Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 9 Rank-Sum Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 10 Rank-Sum Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 11 Rank-Sum Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 12 Spearman Rank Correlation • Suppose we have a sample of size n of paired data points (x, y). • The data points, (x, y), must be ranked variables. • The Spearman Rank Correlation will tell us if the data pairs have a monotone increasing or a monotone decreasing relationship. Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 13 Spearman Rank Correlation Coefficient Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 14 Spearman Rank Correlation Properties Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 15 Spearman Rank Correlation Properties Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 16 Test for Spearman Correlation Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 17 Test for Spearman Correlation Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 18 Spearman Correlation Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 19 Spearman Correlation Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 20 Spearman Correlation Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 21 Runs Test for Randomness • Definitions: Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 22 Runs Test for Randomness Hypotheses Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 23 Conducting the Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 24 Constructing a Runs Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 25 Constructing a Runs Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 26 Constructing a Runs Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 27 Constructing a Runs Test Copyright © Houghton Mifflin Harcourt Publishing Company. All rights reserved. 12 | 28