Other Analysis of Variance Designs Chapter 15. Chapter Topics Basic Experimental Design Concepts  Defining Experimental Design  Controlling Nuisance.

Slides:



Advertisements
Similar presentations
Randomized Complete Block and Repeated Measures (Each Subject Receives Each Treatment) Designs KNNL – Chapters 21,
Advertisements

1 Chapter 4 Experiments with Blocking Factors The Randomized Complete Block Design Nuisance factor: a design factor that probably has an effect.
Chapter 4 Randomized Blocks, Latin Squares, and Related Designs
Analysis of variance (ANOVA)-the General Linear Model (GLM)
Design of Experiments and Analysis of Variance
PSY 307 – Statistics for the Behavioral Sciences
T-Tests.
One-Way Between Subjects ANOVA. Overview Purpose How is the Variance Analyzed? Assumptions Effect Size.
1 Multifactor ANOVA. 2 What We Will Learn Two-factor ANOVA K ij =1 Two-factor ANOVA K ij =1 –Interaction –Tukey’s with multiple comparisons –Concept of.
Chapter 11 Analysis of Variance
Analysis of Variance. Experimental Design u Investigator controls one or more independent variables –Called treatment variables or factors –Contain two.
Chapter 3 Analysis of Variance
Statistics for Managers Using Microsoft® Excel 5th Edition
Lecture 9: One Way ANOVA Between Subjects
8. ANALYSIS OF VARIANCE 8.1 Elements of a Designed Experiment
Chapter 11 Analysis of Variance
Today Concepts underlying inferential statistics
Chapter 14 Inferential Data Analysis
Statistics for Managers Using Microsoft Excel, 4e © 2004 Prentice-Hall, Inc. Chap 13-1 Chapter 13 Introduction to Multiple Regression Statistics for Managers.
Chapter 12 Inferential Statistics Gay, Mills, and Airasian
ANOVA Chapter 12.
QNT 531 Advanced Problems in Statistics and Research Methods
1 of 46 MGMT 6970 PSYCHOMETRICS © 2014, Michael Kalsher Michael J. Kalsher Department of Cognitive Science Inferential Statistics IV: Factorial ANOVA.
Which Test Do I Use? Statistics for Two Group Experiments The Chi Square Test The t Test Analyzing Multiple Groups and Factorial Experiments Analysis of.
1 1 Slide © 2008 Thomson South-Western. All Rights Reserved Chapter 15 Multiple Regression n Multiple Regression Model n Least Squares Method n Multiple.
January 31 and February 3,  Some formulae are presented in this lecture to provide the general mathematical background to the topic or to demonstrate.
t(ea) for Two: Test between the Means of Different Groups When you want to know if there is a ‘difference’ between the two groups in the mean Use “t-test”.
© Copyright McGraw-Hill CHAPTER 12 Analysis of Variance (ANOVA)
PSY 307 – Statistics for the Behavioral Sciences Chapter 16 – One-Factor Analysis of Variance (ANOVA)
Chapter 10 Analysis of Variance.
Psychology 301 Chapters & Differences Between Two Means Introduction to Analysis of Variance Multiple Comparisons.
Randomized Block Design (Kirk, chapter 7) BUSI 6480 Lecture 6.
Testing Hypotheses about Differences among Several Means.
Statistics for Managers Using Microsoft Excel, 4e © 2004 Prentice-Hall, Inc. Chap 10-1 Chapter 10 Analysis of Variance Statistics for Managers Using Microsoft.
Chapter 10: Analyzing Experimental Data Inferential statistics are used to determine whether the independent variable had an effect on the dependent variance.
1 Chapter 15 Introduction to the Analysis of Variance IThe Omnibus Null Hypothesis H 0 :  1 =  2 =... =  p H 1 :  j =  j′ for some j and j´
Educational Research Chapter 13 Inferential Statistics Gay, Mills, and Airasian 10 th Edition.
Chapter 13 Multiple Regression
ANOVA: Analysis of Variance.
Chapter 14 Repeated Measures and Two Factor Analysis of Variance
Analysis of Variance (One Factor). ANOVA Analysis of Variance Tests whether differences exist among population means categorized by only one factor or.
Chapter Seventeen. Figure 17.1 Relationship of Hypothesis Testing Related to Differences to the Previous Chapter and the Marketing Research Process Focus.
Copyright © 2013, 2009, and 2007, Pearson Education, Inc. Chapter 14 Comparing Groups: Analysis of Variance Methods Section 14.3 Two-Way ANOVA.
Chapter 10 The t Test for Two Independent Samples
Chapter 12 Introduction to Analysis of Variance PowerPoint Lecture Slides Essentials of Statistics for the Behavioral Sciences Eighth Edition by Frederick.
Chapter 13 Repeated-Measures and Two-Factor Analysis of Variance
Chapter 11 Analysis of Variance. 11.1: The Completely Randomized Design: One-Way Analysis of Variance vocabulary –completely randomized –groups –factors.
T Test for Two Independent Samples. t test for two independent samples Basic Assumptions Independent samples are not paired with other observations Null.
Chapter 4 Analysis of Variance
IE241: Introduction to Design of Experiments. Last term we talked about testing the difference between two independent means. For means from a normal.
Statistics for Managers Using Microsoft Excel, 5e © 2008 Prentice-Hall, Inc.Chap 14-1 Statistics for Managers Using Microsoft® Excel 5th Edition Chapter.
HYPOTHESIS TESTING FOR DIFFERENCES BETWEEN MEANS AND BETWEEN PROPORTIONS.
ENGR 610 Applied Statistics Fall Week 8 Marshall University CITE Jack Smith.
Factorial BG ANOVA Psy 420 Ainsworth. Topics in Factorial Designs Factorial? Crossing and Nesting Assumptions Analysis Traditional and Regression Approaches.
Educational Research Inferential Statistics Chapter th Chapter 12- 8th Gay and Airasian.
Chapter 12 Introduction to Analysis of Variance
Chapter 14 Repeated Measures and Two Factor Analysis of Variance PowerPoint Lecture Slides Essentials of Statistics for the Behavioral Sciences Seventh.
The 2 nd to last topic this year!!.  ANOVA Testing is similar to a “two sample t- test except” that it compares more than two samples to one another.
Dependent-Samples t-Test
Lecture Slides Elementary Statistics Twelfth Edition
Factorial Experiments
i) Two way ANOVA without replication
Applied Business Statistics, 7th ed. by Ken Black
Comparing Three or More Means
Chapter 11 Analysis of Variance
Other Analysis of Variance Designs
Analysis of Variance (ANOVA)
Randomized Complete Block and Repeated Measures (Each Subject Receives Each Treatment) Designs KNNL – Chapters 21,
STATISTICS INFORMED DECISIONS USING DATA
Presentation transcript:

Other Analysis of Variance Designs Chapter 15

Chapter Topics Basic Experimental Design Concepts  Defining Experimental Design  Controlling Nuisance Variation The Randomized Block and Completely Randomized Factorial Designs  Model Equation  Computational Formulae and Procedures  Assumptions  Interactions  Multiple Comparison Procedures  Practical Significance

Experimental Design Concepts Defining Experimental Design  “Refers to a randomization plan for assigning participants to experimental conditions and the statistical analysis associated with the plan”  We are going to discuss two designs: Randomized Block Design One treatment with two or more levels Uses dependent samples Usually more powerful than a Completely Randomized Design Completely Randomized Factorial Design Two or more treatments, each with two or more levels Nuisance Variation  Any variation in the dependent variable systematically attributable to sources other than the independent variable.

Controlling Nuisance Variation Three experimental approaches to controlling  Hold the nuisance variable constant  Use random sampling / random assignment  Include the nuisance variable as a variable in the experiment Involves forming blocks of participants Participants within a block are more homogeneous with respect to the nuisance variable than those in different blocks Criteria for Blocking  Any variable correlated with the dependent variable other than the independent variable is a candidate to be a blocking variable  Four procedures Repeated Measures Participant Matching Identical twins / littermates Blocks matched by mutual selection

The Randomized Block Design Linear Model Equation for a Score  Given by  The above parameters are generally unknown, but they can be estimated from sample data as follows: The Grand Mean The Treatment Effect The Error Effect The Block Effect

The Randomized Block Design Interpreting the parameters  is the score for the person in the block and the population  is the grand mean of all scores in an experiment  is the effect of the population on this subject’s score  is the effect of the block on this subject’s score  is the random error effect on this subject’s score (continued)

The Randomized Block Design In the RB-p design, we have two different omnibus null hypotheses we can test.  Equality of population treatment means  Equality of population block means (continued)

The Randomized Block Design For this design, we are going to have four terms for sums of squares:  SSTO Sums of Squares Total Measures total variability among scores in an experiment  SSA Sums of Squares Between Treatments Measures variability between treatment levels  SSBL Sums of Squares Within Blocks Measures variability within blocks  SSRES Sums of Squares Residual Measure of “error” variability (continued)

The Randomized Block Design (continued)

The Randomized Block Design This is just another name for the total sample variance! This is the average variability between blocks. This is the average variability between treatment levels. (continued) This is the average “error” variability.

The Randomized Block Design Just like in the CR-p design, in the RB-p design, what we want as researchers is for our “treatment levels” to account for more variation than does random error.  This design (and all designs) also uses the F statistic  Recall from previous chapters that the F statistic is defined as the ratio of two independent variances.  MSA and MSRES are both measures of variance, and they are independent of one another as are MSBL and MSRES.  Forming a ratio for each of these pairs yields two F statistics. (continued)

Computations in an RB-4 Design We have computational formula for all of the sums of squares, and they are given by:

Computations in an RB-4 Design Treatment Levels a1a1 a2a2 a3a3 a4a Using the same data as we used in the CR-p design but modified to include blocks, we will compute these numbers (continued) Treatment Levels a1a1 a2a2 a3a3 a4a4 s1s s2s s3s s4s s5s s6s s7s s8s

Computations in an RB-4 Design Treatment Levels a1a1 a2a2 a3a3 a4a4 s1s s2s s3s s4s s5s s6s s7s s8s (continued) We compute the sums of squares as follows:

Computations in an RB-4 Design (continued) We use these numbers to begin our ANOVA table SourceSSdfMSF Treatment49p-1= ** Blocks32n-1= * Residual29(p-1)(n-1)= Total110np-1=31 **p<.0001 *p<.05

Computations in an RB-4 Design F 3,21 0 This tests the hypothesis that the population treatment means are equal If our computed F exceeds this number, we reject the null hypothesis.  It does, so we reject the null. If the p-value is less than alpha, we reject the null  It is, so we reject Reject  F.05, 3, 21 =3.07 F = (continued)

This tests the hypotheses that the population means for the blocks are equal  Procedure is exactly the same as for testing the other hypothesis, but our df has changed! Computations in an RB-4 Design F 7,21 0 Reject  F.05, 7, 21 =2.49 (continued) F = 3.31

Computations in an RB-4 Design Analyze | Fit Model | Measurement in Y box (Continuous) | Grouping Variable AND Blocking Variable in “Effects” Box (Nominal) |  Different JMP procedure!  Output will be a little different than we’re used to... (continued)

Issues with an RB-p Design In this example, we rejected both omnibus null hypotheses The RB-p design can be more powerful than the CR-p design if... ... we selected a useful nuisance variable and matched participants effectively  What does “effectively” mean? Nuisance variation should account for an appreciable portion of previously unexplained variation The RB-p design can be less powerful than the CR-p design if... ... the loss in error df is not offset by a substantial decrease in error sums of squares

More on the RB-p Design Multiple Comparison Procedures  These are the same as in a CR-p design with one exception: MSRES is used instead of MSWG Practical Significance  Similar measures, but in an RB-p design, we need a “partial” measure  Partial Omega-Squared Measures the strength of association between the dependent variable and the treatment level Values are interpreted the same as before  Cohen’s g

More on the RB-p Design Assumptions  The model equation reflects all the sources of variation that affect each score  The blocks represent a random sample from the population of blocks.  The block effects are normally distributed and the variance of each block population is equal  The block effects are independent of each other and of other effects in the model equation  **The population variances of differences for all pairs of treatment levels are homogeneous  The error effects are normally distributed, and the variances of the error effects are homogeneous  The error effects are independent of one another and of other effects in the model equation (continued)

The CRF-pq Design Factorial Design – two or more treatments Treatment Combination – the combinations of conditions to which a participant is simultaneously exposed  Completely Crossed – each level of one treatment occurs with each level of another treatment  Partially Crossed – each level of one treatment occurs only with some levels of another treatment Linear Model Equation

The CRF-pq Design The Grand Mean Treatment A Effect The Error Effect Treatment B Effect AB Interaction Effect (continued)

The CRF-pq Design In the CRF-pq design, we have three different omnibus null hypotheses we can test.  Equality of population means for treatment A  Equality of population means for treatment B  Treatments A and B do not interact (continued) Interactions equal zero

Computations in a CRF-22 Design Problem: Recall the PPA Example. Restaurant management is interested in determining whether average PPA differs across whether or not patrons consume alcohol in addition to whether or not it differs across gender. There are two treatment levels here  Whether or not patrons consume alcohol – Two levels  Gender – Two levels (Male and Female) First, we need to check to make sure the populations of treatment combinations are symmetric  This is done by obtaining several box-plots, one for each treatment combination, and placing them one above the other

Computations in a CRF-22 Design (continued) We compute the sums of squares as follows:

Computations in a CRF-22 Design (continued) We use these numbers to begin our ANOVA table SourceSSdfMSF Alcohol p-1= ** Gender80.278q-1= * Alcohol X Gender3.5385(p-1)(q-1)= Within Cell pq(n-1)= Total npq-1=31 **p<.0001 *p<.01

Computations in a CRF-22 Design (continued) Interactions  The tests indicate that there is no interaction between gender and whether or not participants consume alcohol  In other words, differences in levels of one treatment are not different at two or more levels of the other treatment  The difference between men and women is essentially the same across ethnicities

Computations in a CRF-22 Design (continued) Interactions  The lines on the previous graph were relatively parallel; this indicates no interaction  If the lines on the graph are non-parallel for some portion of the line, we have interaction.

Computations in a CRF-22 Design F 1,118 0 This tests the hypotheses that the population means for treatment B are equal Procedure is exactly the same for testing the other hypothesis, but our df may change! Reject  F.05, 1, 118 =3.92 (continued) F = 9.78

Practical Significance in a CRF-pq Design Practical Significance  “Partial” measures are used as in the RB-p case.  Partial Omega-Squared Measures the strength of association between the dependent variable and the treatment level For Treatment A... For Treatment B... There is also one for AxB interaction, computed similarly Values are interpreted the same as before  Cohen’s g

CRF-pq Miscellanea Multiple Comparison Procedures are the same as in the CR-p and RB-p designs Assumptions with the CRF-pq Design  The model equation reflects all the sources of variation that affect each score  Participants are random samples from the respective populations or the participants have been randomly assigned to the treatment combinations  The population for each of the treatment combinations is normally distributed  The variance of each of the pq treatment combinations are equal

Chapter Highlights We have developed a definition of experimental design through two designs, each of which are slightly more complex than the CR- p design  Randomized Block Isolates a nuisance variable in hopes of reducing error variance  Completely Randomized Factorial Allows the researcher to test two treatments and the interaction between the treatments Each has the potential to have of greater power than the CR-p design There are many, many more designs out there. We have introduced the three which are most common.