outline review of last week introduction to descriptive statistics the...
TRANSCRIPT
Introduction to ANOVA
2009 Methodology A - Lecture 3
1. Review of Last Week
2. Today’s Learning Objectives
3. What is ANOVA?
4. Types of ANOVA
5. Assumptions
6. Considerations
7. Test of Learning Objectives
8. Vocabulary
Outline Review of Last WeekDescriptive Statistics
1. What are the three most common measures of central tendency?
2. How do you calculate the mean?3. How do you calculate the median?4. How do you calculate the mode?5. What are the two measures of
variability?6. How are these two measures
related?
Hypothesis Testing7. What is the null hypothesis?8. What is the difference between
one-tailed and two-tailed alternative hypotheses?
9. How do p-values relate to the null and alternative hypotheses?
10. How do Type I and Type II errors differ?
The t-test11. What are the three types of t-test?12. What do you need to know about
your data to compute the t statistic?
13. What are the assumptions of the t-test?
14. How do you test for equal variance?
15. What do you do if variance is not equal?
16. What are degrees of freedom?17. How do you calculate effect size?18. How do you report the outcome of
a t-test?19. Given sample data, which type of
t-test is most appropriate?
What is ANOVA
1. What does ANOVA stand for?
2. How is ANOVA similar to a t-test?
3. How is it different?
4. What is a factor?
Types of ANOVA
5. What is the difference between
univariate and multivariate ANOVAs?
6. What is the difference between
between-subjects and within-subject
factors?
7. What is the difference between one-
way and factorial ANOVAs?
8. For a univariate design, what 2 things
do you need to know to determine
what type of ANOVA to use?
9. What type of ANOVA is required if
you have both between-subjects and
within-subject factors?
Assumptions
10. What are the three main assumptions
of ANOVA?
11. What descriptive statistics do you
report to assess normality?
12. What are the two tests for
homogeneity of variance?
13. When should you use each of the
tests for homogeneity of variance?
14. How do you compute Fmax?
15. When do you need to check for
sphericity?
16. What values of Levene’s Test, Fmax
and Mauchly’s Test allow you to do
ANOVA?
Other Considerations
17. Why should you consider sample size
when planning an experiment?
18. What is meant by ‘cases must be
independent’?
Today’s Learning Objectives! ANalysis Of VAriance
! Like a t-test for 2 or more conditionsfor 2 conditions, F = t2
! Also used for multiple factors (>1 independent variable)
! A parametric statistic (has assumptions)
What is ANOVA?
Number of Independent Variables (IV)
! One-way - 1 factor
! Factorial - 2 or more orthogonal factors
Types of ANOVA
Groups of Subjects
! Between-subjects - 2 or more groups of subjects, each subject participates in 1 condition
! Within-subjects - 1 group of subjects, each subject participates in all conditions
Number of Dependent Variables (DV)
! Univariate - 1 DV! Repeated-measures - 1 DV measured 2 or more times
! Multivariate - 2 or more different DVs
Types of ANOVA
One IV More than one IV
OneOne-way between-subjects
Factorialbetween-subjects
Mixed-design(split-plot)
AllOne-way
within-subjectFactorial
within-subject
Number of Independent Variables
Conditio
ns p
er
Subje
ct
plus 1 or more continuous IVs = ANCOVA
1. The sample is drawn from a normally-distributed population
2. Homogeneity of variance
3. Sphericity (only for within-subjects designs)
Assumptions
155-6
Always look at your data first. Remove outliers and “eyeball” for normality.
SPSS ! Graphs ! Chart Builder...
1. Normal Distribution
1. Normal DistributionThere are many ways to determine if data are normally distributed and ANOVA is robust to most violations of normality. For this course, assume data meet the assumptions and just report skewness and kurtosis
platykurtic mesokurtic leptokurtic
negative skew zero skew positive skew
1. Normal DistributionReporting skewness and kurtosis
1. Normal DistributionReporting skewness and kurtosis
Test that the variance of each condition is roughly equal using Levene’s Test for between-subjects factors and Fmax for within-subject factors.
2. Homogeneity of Variance
equal variance unequal variance
Test that the variance of each condition is roughly equal using Levene’s Test for between-subjects factors and Fmax for within-subject factors.
If p > .05, variance is equal enough for ANOVA
2. Homogeneity of VarianceTest that the variance of each condition is roughly equal using Levene’s Test for between-subjects factors and Fmax for within-subject factors.
Fmax =largest variance
smallest variance Fmax = = 1.6389.783
55.201
If Fmax < 4, variance is equal enough for ANOVA
2. Homogeneity of Variance
3. Sphericity! For within-subject factors with more than 2 levels, you must
check for and report sphericity.
! Sphericity is like homogeneity of variance for difference scores (the difference between pairs of within-subject factors).
! SPSS does this automatically via Mauchly’s Test of Sphericity.
! If p > 0.05, report the ‘Sphericity Assumed’ statistics, else, report the Greenhouse-Geisser statistics.
1. Cases must be independent
2. Sample size should be approximately equal for each group
3. Samples should not be too small
Considerations when designing experiments to be analysed by ANOVA
ANCOVAANOVAbetween-subejctsdependent variable (DV)Fmax
factorfactorialGreenhouse-Geiserhomogeneity of variancehomoscedasticitykurtosisLevene statisticMANOVA
Mauchly’s testmixed-designmultivariateindependent variable (IV)one-wayorthogonalrepeated-measuresrobustskewnesssphericitysplit-plot ANOVAunivariatewithin-subject
Vocabulary