Experimental Design Tutorial Presented By Michael W. Totaro Wireless Research Group Center for Advanced Computer Studies University of Louisiana at Lafayette.

Slides:



Advertisements
Similar presentations
Analysis by design Statistics is involved in the analysis of data generated from an experiment. It is essential to spend time and effort in advance to.
Advertisements

Chapter 6 The 2k Factorial Design
LSU-HSC School of Public Health Biostatistics 1 Statistical Core Didactic Introduction to Biostatistics Donald E. Mercante, PhD.
Variance reduction techniques. 2 Introduction Simulation models should be coded such that they are efficient. Efficiency in terms of programming ensures.
Experimental Design, Response Surface Analysis, and Optimization
Introduction to One way and Two Way analysis of Variance......
1 PERFORMANCE EVALUATION H Often one needs to design and conduct an experiment in order to: – demonstrate that a new technique or concept is feasible –demonstrate.
Statistics CSE 807.
1/55 EF 507 QUANTITATIVE METHODS FOR ECONOMICS AND FINANCE FALL 2008 Chapter 10 Hypothesis Testing.
Evaluating Hypotheses
Chapter 9 Audit Sampling: An Application to Substantive Tests of Account Balances McGraw-Hill/Irwin ©2008 The McGraw-Hill Companies, All Rights Reserved.
1 PERFORMANCE EVALUATION H Often in Computer Science you need to: – demonstrate that a new concept, technique, or algorithm is feasible –demonstrate that.
IE 429 DESIGNS OF EXPERIMENTS
Basic Business Statistics, 10e © 2006 Prentice-Hall, Inc. Chap 9-1 Chapter 9 Fundamentals of Hypothesis Testing: One-Sample Tests Basic Business Statistics.
Lecture 10 Comparison and Evaluation of Alternative System Designs.
Inferences About Process Quality
Lehrstuhl für Informatik 2 Gabriella Kókai: Maschine Learning 1 Evaluating Hypotheses.
Statistics for Managers Using Microsoft® Excel 5th Edition
Chapter 7 Probability and Samples: The Distribution of Sample Means
Introduction to the design (and analysis) of experiments James M. Curran Department of Statistics, University of Auckland
T WO WAY ANOVA WITH REPLICATION  Also called a Factorial Experiment.  Replication means an independent repeat of each factor combination.  The purpose.
Objectives of Multiple Regression
1 1 Slide © 2003 South-Western/Thomson Learning™ Slides Prepared by JOHN S. LOUCKS St. Edward’s University.
Inference for regression - Simple linear regression
Chapter 10 Hypothesis Testing
Business Statistics, A First Course (4e) © 2006 Prentice-Hall, Inc. Chap 9-1 Chapter 9 Fundamentals of Hypothesis Testing: One-Sample Tests Business Statistics,
Fundamentals of Hypothesis Testing: One-Sample Tests
Introduction to Statistical Inferences
AZIZ KUSTIYO DEPARTEMEN ILMU KOMPUTER FMIPA IPB. INTRODUCTION TO EXPERIMENTAL DESIGN  The goal of a proper experimental design is to obtain the maximum.
Chapter 1: Introduction to Statistics
QNT 531 Advanced Problems in Statistics and Research Methods
Introduction ANOVA Mike Tucker School of Psychology B209 Portland Square University of Plymouth Drake Circus Plymouth, PL4 8AA Tel: +44 (0)
QBM117 Business Statistics Estimating the population mean , when the population variance  2, is known.
Chapter 10 Hypothesis Testing
Analysis of the Impact and Interactions of Protocol and Environmental Parameters on Overall MANET Performance Michael W. Totaro and Dmitri D. Perkins Center.
© 1998, Geoff Kuenning General 2 k Factorial Designs Used to explain the effects of k factors, each with two alternatives or levels 2 2 factorial designs.
MS 305 Recitation 11 Output Analysis I
Factorial Design of Experiments Kevin Leyton-Brown.
Introduction to Experimental Design
ANOVA: Analysis of Variance
Yaomin Jin Design of Experiments Morris Method.
Copyright © 2006 by The McGraw-Hill Companies, Inc. All rights reserved. McGraw-Hill/Irwin 9-1 Chapter Nine Audit Sampling: An Application to Substantive.
Measures of central tendency are statistics that express the most typical or average scores in a distribution These measures are: The Mode The Median.
Chapter 4 Linear Regression 1. Introduction Managerial decisions are often based on the relationship between two or more variables. For example, after.
Chapter 7 Probability and Samples: The Distribution of Sample Means
Statistics for Managers Using Microsoft Excel, 4e © 2004 Prentice-Hall, Inc. Chap 8-1 Chapter 8 Fundamentals of Hypothesis Testing: One-Sample Tests Statistics.
Lecture 9 Chap 9-1 Chapter 2b Fundamentals of Hypothesis Testing: One-Sample Tests.
1 Doing Statistics for Business Doing Statistics for Business Data, Inference, and Decision Making Marilyn K. Pelosi Theresa M. Sandifer Chapter 14 Experimental.
Statistics : Statistical Inference Krishna.V.Palem Kenneth and Audrey Kennedy Professor of Computing Department of Computer Science, Rice University 1.
Copyright © 2013, 2009, and 2007, Pearson Education, Inc. Chapter 14 Comparing Groups: Analysis of Variance Methods Section 14.3 Two-Way ANOVA.
Copyright © Cengage Learning. All rights reserved. 12 Analysis of Variance.
Math 4030 – 9a Introduction to Hypothesis Testing
Chapter 3 Data Description Section 3-2 Measures of Central Tendency.
© Copyright McGraw-Hill 2004
Tutorial I: Missing Value Analysis
1 Simulation Scenarios. 2 Computer Based Experiments Systematically planning and conducting scientific studies that change experimental variables together.
1 Design of Engineering Experiments – The 2 k Factorial Design Text reference, Chapter 6 Special case of the general factorial design; k factors, all at.
Designs for Experiments with More Than One Factor When the experimenter is interested in the effect of multiple factors on a response a factorial design.
McGraw-Hill/IrwinCopyright © 2009 by The McGraw-Hill Companies, Inc. All Rights Reserved. Experimental Design and Analysis of Variance Chapter 11.
Chapter 11: Test for Comparing Group Means: Part I.
Chapter 9 Introduction to the t Statistic
Virtual University of Pakistan
Stats Methods at IC Lecture 3: Regression.
Statistical Core Didactic
Hypothesis Testing and Confidence Intervals (Part 1): Using the Standard Normal Lecture 8 Justin Kern October 10 and 12, 2017.
Counting Statistics and Error Prediction
Inferential Statistics
Advanced Algebra Unit 1 Vocabulary
DESIGN OF EXPERIMENTS by R. C. Baker
Presumptions Subgroups (samples) of data are formed.
Presentation transcript:

Experimental Design Tutorial Presented By Michael W. Totaro Wireless Research Group Center for Advanced Computer Studies University of Louisiana at Lafayette

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Introduction  Broad goal of simulation projects is to learn how the inputs affect the outputs  Kinds of factors (input parameters) Quantitative vs. Qualitative Controllable vs. Uncontrollable  In modeling, everything is controllable  Simulation output performance measures are the responses

Analogy to Traditional Physical Experiments Laboratory Agricultural Industrial

Goal  In simulation, experimental design provides a way of deciding before the runs are made which particular configurations to simulate so that the desired information can be obtained with the least amount of simulating.

Possible Factors/Responses Usually, there are many possible factors and responses

Setting Factor Levels  There is no real prescription for setting factor levels (i.e., values they can take on) Qualitative—may be clear from context Quantitative—may set at “reasonable” levels; however, that might push the boundaries

Opportunities  Special opportunities in simulation- based experiment Everything is controllable Control source of randomness, and exploit for variance reduction No need to randomize assignment of treatments to experimental results

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Feasible Design  Example of a design that is feasible in many simulations: 2 k factorial design  Have k factors (inputs), each at just two levels  Number of possible combinations of factors—usually called design points— is 2 k

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Single Factor vs. Multiple Factors  Case of single factor (k = 1) Vary the factor (maybe at more than two levels), make plots, and so on  In general, assume k ≥ 2 factors— want to know about: Effect on response(s) of each factor Possible interactions between factors— effect of one factor depends on the level of some of the other factors

2 k Factorial Design—Process  Code each factor to a “+” and a “-” level  Design matrix: All possible combinations of factor levels  Example for k = 3 factors: Make the 8 simulation runs, and measure the effects of the factors!

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Main Effect of a Factor Main effect of a factor is the average difference in the response when this factor is at its “+” level as opposed to its “-” level:

Main Effect of a Factor – cont’d The main effects measure the average change in the response due to a change in an individual factor, with this average being taken over all possible combinations of the other k-1 factors (numbering 2 k-1 ).

Main Effect of a Factor – cont’d We can rewrite the above as “Factor 1” column ● “Response” column / 2 k-1 -R 1 + R 2 – R 3 + R 4 – R 5 + R 6 – R 7 + R 8 e 1 = 4

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Factor Interaction  Two factors A and B are said to interact if the effect of one depends upon the level of the other  Conversely, these two factors, A and B, are said to be noninteracting if the performance of one is not affected by the level of the other  We shall look at examples of interacting factors and noninteracting factors

Examples of Noninteracting and Interacting Factors A1A1 A2A2 B1B1 35 B2B2 68 Noninteracting Factors Interacting Factors A1A1 A2A2 B1B1 35 B2B2 69 As the factor A is changed from level A 1 to level A 2, the performance increases by 2 regardless of the level of factor B As the factor A is changed from level A 1 to level A 2, the performance increases either by 2 or 3 depending upon whether B is at level B 1 or level B 2, respectively

Examples of Noninteracting and Interacting Factors—cont’d Performance Graphical representation of interacting and noninteracting factors A1A1 A2A2 B2B2 B1B1 Performance B1B1 A2A2 A1A1 B2B2 (a) No Interaction Performance A1A1 A2A2 B2B2 B1B B1B1 A2A2 A1A1 B2B2 (b) Interaction

Interaction Effects 1 x 3 interaction effect: “Factor 1” ● “Factor 3” ● “Response” / 2 k-1 R 1 - R 2 + R 3 - R 4 – R 5 + R 6 – R 7 + R 8 e 13 = 4  Addresses the question: “Does the effect of a factor depend on level of others?”  Sign of effect indicates direction of effect on response of moving that factor from its “-” to its “+” level

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Quantifying the Effects  Statistical significance of effects estimates (i.e., are they real?)  A luxury in simulation-based experiments: Replicate the whole design n times Get n observations on each effect Compute sample mean, sample variance, confidence interval, etc., on expected effects—effect is “real” if confidence interval misses 0

Quantifying the Effects--Example  Example of 2 6 Factorial Design  In addition to above, machine suffers breakdowns, and thus must undergo repair  Response: Average time in system of a part (called the makespan)

Quantifying the Effects—Example (cont’d)  Factors and coding:  Full 2 6 factorial design involves 64 factor combinations  Entire design is replicated n = 5 times; thus, this is a factorial experimental design

Quantifying the Effects—Example (cont’d) The figures below plot 90% confidence intervals of the expected main effects and two-way way interactions for both responses, obtained by the five replications of the entire design We see that factor 2 (inspection time) has a large negative effect on makespan—thus, “improving” it to “+” level would be the single most worthwhile step to take to reduce makespan. (Put another way, faster inspections would provide the greatest improvement.) Improving factor 5 (probability of failing inspection) would have the next-most-important effect on makespan

Topics  Introduction  2 k Factorial Designs  Factors/Responses  Effects  Factor Interaction  Quantifying the Effects  Proper Perspective

Keep a Proper Perspective  Results are relative to the particular values chosen for the factors, and cannot necessarily be extrapolated to other regions in the factor space  It is probably not good to choose the “-” and “+” levels of a factor to be extremely far apart from each other Could result in experiments for factor levels that are unrealistic in the problem context Might not get information on “interior” of design space between the factor levels; thus, we may not see interactions that might be present there

Sources  Simulation Modeling and Analysis, Third Ed., by Averill M. Law and W. David Kelton, The McGraw-Hill Companies, Inc.,  The Art of Computer Systems Performance Analysis: Techniques for Experimental Design, Measurement, Simulation, and Modeling, by Raj Jain, John Wiley & Sons, Inc., New York, 1991.