analysis of variance (anova) - james · pdf fileone-way anova two-way anova analysis of...
Post on 15-Mar-2018
299 Views
Preview:
TRANSCRIPT
One-Way ANOVATwo-way ANOVA
Analysis of Variance (ANOVA)
MGMT 662: Integrative Research Project
August 7, 2008.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Goals
Goals of this class meeting 2/ 15
Learn how to test for significant differences in means from twoor more groups.
Learn how to account for an additional factor.
Learn how to test for significant differences in medians fromtwo or more groups. Why?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
One-Way ANOVA 3/ 15
Method for testing for significant differences among meansfrom two or more groups.
Essentially an extension of the t-test for testing the differencesbetween two means.
Uses measures of variance to measure for differences in means.
Total variation in your data is decomposed into twocomponents:
Among-group variation: variability that is due to differencesamong groups, also called explained variation.Within-group variation: total variability within each of thegroups, this is unexplained variation.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
One-Way ANOVA 3/ 15
Method for testing for significant differences among meansfrom two or more groups.
Essentially an extension of the t-test for testing the differencesbetween two means.
Uses measures of variance to measure for differences in means.
Total variation in your data is decomposed into twocomponents:
Among-group variation: variability that is due to differencesamong groups, also called explained variation.Within-group variation: total variability within each of thegroups, this is unexplained variation.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
One-Way ANOVA 3/ 15
Method for testing for significant differences among meansfrom two or more groups.
Essentially an extension of the t-test for testing the differencesbetween two means.
Uses measures of variance to measure for differences in means.
Total variation in your data is decomposed into twocomponents:
Among-group variation: variability that is due to differencesamong groups, also called explained variation.Within-group variation: total variability within each of thegroups, this is unexplained variation.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
One-Way ANOVA 3/ 15
Method for testing for significant differences among meansfrom two or more groups.
Essentially an extension of the t-test for testing the differencesbetween two means.
Uses measures of variance to measure for differences in means.
Total variation in your data is decomposed into twocomponents:
Among-group variation: variability that is due to differencesamong groups, also called explained variation.Within-group variation: total variability within each of thegroups, this is unexplained variation.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
One-Way ANOVA 3/ 15
Method for testing for significant differences among meansfrom two or more groups.
Essentially an extension of the t-test for testing the differencesbetween two means.
Uses measures of variance to measure for differences in means.
Total variation in your data is decomposed into twocomponents:
Among-group variation: variability that is due to differencesamong groups, also called explained variation.Within-group variation: total variability within each of thegroups, this is unexplained variation.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
One-Way ANOVA 3/ 15
Method for testing for significant differences among meansfrom two or more groups.
Essentially an extension of the t-test for testing the differencesbetween two means.
Uses measures of variance to measure for differences in means.
Total variation in your data is decomposed into twocomponents:
Among-group variation: variability that is due to differencesamong groups, also called explained variation.Within-group variation: total variability within each of thegroups, this is unexplained variation.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Variance Decomposition 4/ 15
Sum of squares groups (SSG):
SSG =K∑
k=1
nk (x̄k − x̄)2
K: number of groups.x̄k is the mean of group k.x̄ is the mean of all the data.
Sum of squares within-group (SSW):
SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄k)2
Sum of squares total (SST):
SST = SSG + SSW =K∑
k=1
nk∑i=1
(xi ,k − x̄)2
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Hypothesis Test 5/ 15
Null hypothesis: µ1 = µ2 = ... = µK
Alternative hypothesis: At least one of the means are differentfrom the others.
F-test (has an F-distribution with degrees of freedom K − 1,n − 1):
F =SSG/(K − 1)
SSW /(n − 1)
Intuitively, what is implied when the F-statistic is large?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Hypothesis Test 5/ 15
Null hypothesis: µ1 = µ2 = ... = µK
Alternative hypothesis: At least one of the means are differentfrom the others.
F-test (has an F-distribution with degrees of freedom K − 1,n − 1):
F =SSG/(K − 1)
SSW /(n − 1)
Intuitively, what is implied when the F-statistic is large?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Hypothesis Test 5/ 15
Null hypothesis: µ1 = µ2 = ... = µK
Alternative hypothesis: At least one of the means are differentfrom the others.
F-test (has an F-distribution with degrees of freedom K − 1,n − 1):
F =SSG/(K − 1)
SSW /(n − 1)
Intuitively, what is implied when the F-statistic is large?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Hypothesis Test 5/ 15
Null hypothesis: µ1 = µ2 = ... = µK
Alternative hypothesis: At least one of the means are differentfrom the others.
F-test (has an F-distribution with degrees of freedom K − 1,n − 1):
F =SSG/(K − 1)
SSW /(n − 1)
Intuitively, what is implied when the F-statistic is large?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Hypothesis Test 5/ 15
Null hypothesis: µ1 = µ2 = ... = µK
Alternative hypothesis: At least one of the means are differentfrom the others.
F-test (has an F-distribution with degrees of freedom K − 1,n − 1):
F =SSG/(K − 1)
SSW /(n − 1)
Intuitively, what is implied when the F-statistic is large?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions behind One-way ANOVA F-test 6/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Sufficiently large (?) sample size, or else population musthave a normal distribution.
Homogeneity of variance: the variances of each of the Kgroups must be equal (σ2
1 = σ22 = ...σ2
K ).
Levene test for homogeneity of variance can be used to test forthis.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions behind One-way ANOVA F-test 6/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Sufficiently large (?) sample size, or else population musthave a normal distribution.
Homogeneity of variance: the variances of each of the Kgroups must be equal (σ2
1 = σ22 = ...σ2
K ).
Levene test for homogeneity of variance can be used to test forthis.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions behind One-way ANOVA F-test 6/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Sufficiently large (?) sample size, or else population musthave a normal distribution.
Homogeneity of variance: the variances of each of the Kgroups must be equal (σ2
1 = σ22 = ...σ2
K ).
Levene test for homogeneity of variance can be used to test forthis.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions behind One-way ANOVA F-test 6/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Sufficiently large (?) sample size, or else population musthave a normal distribution.
Homogeneity of variance: the variances of each of the Kgroups must be equal (σ2
1 = σ22 = ...σ2
K ).
Levene test for homogeneity of variance can be used to test forthis.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions behind One-way ANOVA F-test 6/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Sufficiently large (?) sample size, or else population musthave a normal distribution.
Homogeneity of variance: the variances of each of the Kgroups must be equal (σ2
1 = σ22 = ...σ2
K ).
Levene test for homogeneity of variance can be used to test forthis.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Example: Crime Rates 7/ 15
Data on 47 states from 1960 (I know its old) on the crime rateand a number of factors that may influence the crime rate.In particular, I made a variable that put unemployment intocategories:
Unemployment = 1 if unemployment rate was less than 8%.Unemployment = 2 if unemployment rate was between 8 and10%.Unemployment = 3 if unemployment rate was greater than10%.
I also made a variable that categorized schooling:Schooling = 1 if mean years of schooling for given state wasless than 10 years.Schooling = 2 otherwise.
Is there statistical evidence that the mean crime rate isdifferent among the different categories for the level ofunemployment?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Example: Crime Rates 7/ 15
Data on 47 states from 1960 (I know its old) on the crime rateand a number of factors that may influence the crime rate.In particular, I made a variable that put unemployment intocategories:
Unemployment = 1 if unemployment rate was less than 8%.Unemployment = 2 if unemployment rate was between 8 and10%.Unemployment = 3 if unemployment rate was greater than10%.
I also made a variable that categorized schooling:Schooling = 1 if mean years of schooling for given state wasless than 10 years.Schooling = 2 otherwise.
Is there statistical evidence that the mean crime rate isdifferent among the different categories for the level ofunemployment?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Example: Crime Rates 7/ 15
Data on 47 states from 1960 (I know its old) on the crime rateand a number of factors that may influence the crime rate.In particular, I made a variable that put unemployment intocategories:
Unemployment = 1 if unemployment rate was less than 8%.Unemployment = 2 if unemployment rate was between 8 and10%.Unemployment = 3 if unemployment rate was greater than10%.
I also made a variable that categorized schooling:Schooling = 1 if mean years of schooling for given state wasless than 10 years.Schooling = 2 otherwise.
Is there statistical evidence that the mean crime rate isdifferent among the different categories for the level ofunemployment?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Example: Crime Rates 7/ 15
Data on 47 states from 1960 (I know its old) on the crime rateand a number of factors that may influence the crime rate.In particular, I made a variable that put unemployment intocategories:
Unemployment = 1 if unemployment rate was less than 8%.Unemployment = 2 if unemployment rate was between 8 and10%.Unemployment = 3 if unemployment rate was greater than10%.
I also made a variable that categorized schooling:Schooling = 1 if mean years of schooling for given state wasless than 10 years.Schooling = 2 otherwise.
Is there statistical evidence that the mean crime rate isdifferent among the different categories for the level ofunemployment?
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Nonparametric One-way ANOVA 8/ 15
Kruskal-Wallis Rank Test: non-parametric technique fortesting for differences in the medians among two or moregroups.
Like the Mann-Whitney U-test, uses information about theranks of the observations, instead of the actual sizes.
Null hypothesis: θ1 = θ2 = ... = θK (i.e. all groups have thesame median).
Alternative hypothesis: at least one of the medians differ.
As the sample size gets large (over 5 per group some say!),the Kruskal-Wallis test statistic approaches a χ2 distributionwith K − 1 degrees of freedom.
For small sample sizes: possible to compute exact p-valueswithout depending on asymptotic distributions.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Nonparametric One-way ANOVA 8/ 15
Kruskal-Wallis Rank Test: non-parametric technique fortesting for differences in the medians among two or moregroups.
Like the Mann-Whitney U-test, uses information about theranks of the observations, instead of the actual sizes.
Null hypothesis: θ1 = θ2 = ... = θK (i.e. all groups have thesame median).
Alternative hypothesis: at least one of the medians differ.
As the sample size gets large (over 5 per group some say!),the Kruskal-Wallis test statistic approaches a χ2 distributionwith K − 1 degrees of freedom.
For small sample sizes: possible to compute exact p-valueswithout depending on asymptotic distributions.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Nonparametric One-way ANOVA 8/ 15
Kruskal-Wallis Rank Test: non-parametric technique fortesting for differences in the medians among two or moregroups.
Like the Mann-Whitney U-test, uses information about theranks of the observations, instead of the actual sizes.
Null hypothesis: θ1 = θ2 = ... = θK (i.e. all groups have thesame median).
Alternative hypothesis: at least one of the medians differ.
As the sample size gets large (over 5 per group some say!),the Kruskal-Wallis test statistic approaches a χ2 distributionwith K − 1 degrees of freedom.
For small sample sizes: possible to compute exact p-valueswithout depending on asymptotic distributions.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Nonparametric One-way ANOVA 8/ 15
Kruskal-Wallis Rank Test: non-parametric technique fortesting for differences in the medians among two or moregroups.
Like the Mann-Whitney U-test, uses information about theranks of the observations, instead of the actual sizes.
Null hypothesis: θ1 = θ2 = ... = θK (i.e. all groups have thesame median).
Alternative hypothesis: at least one of the medians differ.
As the sample size gets large (over 5 per group some say!),the Kruskal-Wallis test statistic approaches a χ2 distributionwith K − 1 degrees of freedom.
For small sample sizes: possible to compute exact p-valueswithout depending on asymptotic distributions.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Nonparametric One-way ANOVA 8/ 15
Kruskal-Wallis Rank Test: non-parametric technique fortesting for differences in the medians among two or moregroups.
Like the Mann-Whitney U-test, uses information about theranks of the observations, instead of the actual sizes.
Null hypothesis: θ1 = θ2 = ... = θK (i.e. all groups have thesame median).
Alternative hypothesis: at least one of the medians differ.
As the sample size gets large (over 5 per group some say!),the Kruskal-Wallis test statistic approaches a χ2 distributionwith K − 1 degrees of freedom.
For small sample sizes: possible to compute exact p-valueswithout depending on asymptotic distributions.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Nonparametric One-way ANOVA 8/ 15
Kruskal-Wallis Rank Test: non-parametric technique fortesting for differences in the medians among two or moregroups.
Like the Mann-Whitney U-test, uses information about theranks of the observations, instead of the actual sizes.
Null hypothesis: θ1 = θ2 = ... = θK (i.e. all groups have thesame median).
Alternative hypothesis: at least one of the medians differ.
As the sample size gets large (over 5 per group some say!),the Kruskal-Wallis test statistic approaches a χ2 distributionwith K − 1 degrees of freedom.
For small sample sizes: possible to compute exact p-valueswithout depending on asymptotic distributions.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions for Kruskal-Wallis Test 9/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Only the location (i.e. the center) of the distributions differamong the groups. The populations otherwise have the samedistribution.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions for Kruskal-Wallis Test 9/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Only the location (i.e. the center) of the distributions differamong the groups. The populations otherwise have the samedistribution.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestNonparametric Test
Assumptions for Kruskal-Wallis Test 9/ 15
Randomness: individual observations are assigned to groupsrandomly.
Independence: individuals in each group are independent fromindividuals in another group.
Only the location (i.e. the center) of the distributions differamong the groups. The populations otherwise have the samedistribution.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Two-way ANOVA 10/ 15
One-way ANOVA, the effects of one factor where examined.
Two-way ANOVA, also called two-factor factorial design:two factors are simultaneously evaluated.
Total variance is decomposed into:
variability explained by being in different groups of factor A.variability explained by being in different groups of factor B.variability explained by the interaction of factors A and B.unexplained variability.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
ANOVA Descriptive Statistics 11/ 15
Unemployment LevelSchooling Level Less than 8% 8% to 10% More than 10%
Less than 10 years x̄11 x̄12 x̄13
10 years or more x̄21 x̄22 x̄23
Goals of ANOVA:
Determine if schooling level (factor A) leads to different levelsfor crime rate.Determine if unemployment level (factor B) leads to differentlevels for crime rate.Determine if schooling and unemployment have a joint effecton crime rate (i.e. does the unemployment level effect theimpact of schooling on the crime rate, or vice versa).
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
ANOVA Descriptive Statistics 11/ 15
Unemployment LevelSchooling Level Less than 8% 8% to 10% More than 10%
Less than 10 years x̄11 x̄12 x̄13
10 years or more x̄21 x̄22 x̄23
Goals of ANOVA:
Determine if schooling level (factor A) leads to different levelsfor crime rate.Determine if unemployment level (factor B) leads to differentlevels for crime rate.Determine if schooling and unemployment have a joint effecton crime rate (i.e. does the unemployment level effect theimpact of schooling on the crime rate, or vice versa).
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
ANOVA Descriptive Statistics 11/ 15
Unemployment LevelSchooling Level Less than 8% 8% to 10% More than 10%
Less than 10 years x̄11 x̄12 x̄13
10 years or more x̄21 x̄22 x̄23
Goals of ANOVA:
Determine if schooling level (factor A) leads to different levelsfor crime rate.Determine if unemployment level (factor B) leads to differentlevels for crime rate.Determine if schooling and unemployment have a joint effecton crime rate (i.e. does the unemployment level effect theimpact of schooling on the crime rate, or vice versa).
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
ANOVA Descriptive Statistics 11/ 15
Unemployment LevelSchooling Level Less than 8% 8% to 10% More than 10%
Less than 10 years x̄11 x̄12 x̄13
10 years or more x̄21 x̄22 x̄23
Goals of ANOVA:
Determine if schooling level (factor A) leads to different levelsfor crime rate.Determine if unemployment level (factor B) leads to differentlevels for crime rate.Determine if schooling and unemployment have a joint effecton crime rate (i.e. does the unemployment level effect theimpact of schooling on the crime rate, or vice versa).
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor A 12/ 15
H0 : µ1. = µ2. = ... = µr .
Ha : At least on of the means of the groups in factor A aredifferent from the others.
F =SSA/(r − 1)
SSE/(N − rc)
F-statistic has degrees of freedom r − 1, N − rc .
N =∑r
i=1 ni ..
r (number of rows) is the number of groups for factor A.
c (number of columns) is the number of groups for factor B.
µi . is the mean of group i of factor A.
SSA is the sum of squares from factor A.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor B 13/ 15
H0 : µ.1 = µ.2 = ... = µ.c
Ha : At least on of the means of the groups in factor B aredifferent from the others.
F =SSB/(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom c − 1, N − rc .
µj . is the mean of group j of factor B.
SSB is the sum of squares from factor B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor B 13/ 15
H0 : µ.1 = µ.2 = ... = µ.c
Ha : At least on of the means of the groups in factor B aredifferent from the others.
F =SSB/(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom c − 1, N − rc .
µj . is the mean of group j of factor B.
SSB is the sum of squares from factor B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor B 13/ 15
H0 : µ.1 = µ.2 = ... = µ.c
Ha : At least on of the means of the groups in factor B aredifferent from the others.
F =SSB/(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom c − 1, N − rc .
µj . is the mean of group j of factor B.
SSB is the sum of squares from factor B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor B 13/ 15
H0 : µ.1 = µ.2 = ... = µ.c
Ha : At least on of the means of the groups in factor B aredifferent from the others.
F =SSB/(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom c − 1, N − rc .
µj . is the mean of group j of factor B.
SSB is the sum of squares from factor B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor B 13/ 15
H0 : µ.1 = µ.2 = ... = µ.c
Ha : At least on of the means of the groups in factor B aredifferent from the others.
F =SSB/(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom c − 1, N − rc .
µj . is the mean of group j of factor B.
SSB is the sum of squares from factor B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Factor B 13/ 15
H0 : µ.1 = µ.2 = ... = µ.c
Ha : At least on of the means of the groups in factor B aredifferent from the others.
F =SSB/(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom c − 1, N − rc .
µj . is the mean of group j of factor B.
SSB is the sum of squares from factor B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Interaction of Factors A and B 14/ 15
H0 : there is no interaction effect.
Ha : there is an interaction effect.
F =SSAB/(r − 1)(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom (r − 1)(c − 1), N − rc .
SSAB is the sum of squares from factors A and B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Interaction of Factors A and B 14/ 15
H0 : there is no interaction effect.
Ha : there is an interaction effect.
F =SSAB/(r − 1)(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom (r − 1)(c − 1), N − rc .
SSAB is the sum of squares from factors A and B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Interaction of Factors A and B 14/ 15
H0 : there is no interaction effect.
Ha : there is an interaction effect.
F =SSAB/(r − 1)(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom (r − 1)(c − 1), N − rc .
SSAB is the sum of squares from factors A and B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Interaction of Factors A and B 14/ 15
H0 : there is no interaction effect.
Ha : there is an interaction effect.
F =SSAB/(r − 1)(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom (r − 1)(c − 1), N − rc .
SSAB is the sum of squares from factors A and B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Hypothesis Test for Interaction of Factors A and B 14/ 15
H0 : there is no interaction effect.
Ha : there is an interaction effect.
F =SSAB/(r − 1)(c − 1)
SSE/(N − rc)
F-statistic has degrees of freedom (r − 1)(c − 1), N − rc .
SSAB is the sum of squares from factors A and B.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Nonparametric Two-Way ANOVA 15/ 15
Advise: run two, one-way ANOVA’s using Kruskal-Wallis test.
Or.. run a Kruskal Wallis test with r x c different groups:
Group 1: Unemployment=1, Schooling=1Group 2: Unemployment=2, Schooling=1Group 3: Unemployment=3, Schooling=1Group 4: Unemployment=1, Schooling=2Group 5: Unemployment=2, Schooling=2Group 6: Unemployment=3, Schooling=2
The Tukey pairwise tests for these combinations will indicateif there are interaction effects.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Nonparametric Two-Way ANOVA 15/ 15
Advise: run two, one-way ANOVA’s using Kruskal-Wallis test.
Or.. run a Kruskal Wallis test with r x c different groups:
Group 1: Unemployment=1, Schooling=1Group 2: Unemployment=2, Schooling=1Group 3: Unemployment=3, Schooling=1Group 4: Unemployment=1, Schooling=2Group 5: Unemployment=2, Schooling=2Group 6: Unemployment=3, Schooling=2
The Tukey pairwise tests for these combinations will indicateif there are interaction effects.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
One-Way ANOVATwo-way ANOVA
Variance DecompositionParametric TestsNonparametric Two-way ANOVA
Nonparametric Two-Way ANOVA 15/ 15
Advise: run two, one-way ANOVA’s using Kruskal-Wallis test.
Or.. run a Kruskal Wallis test with r x c different groups:
Group 1: Unemployment=1, Schooling=1Group 2: Unemployment=2, Schooling=1Group 3: Unemployment=3, Schooling=1Group 4: Unemployment=1, Schooling=2Group 5: Unemployment=2, Schooling=2Group 6: Unemployment=3, Schooling=2
The Tukey pairwise tests for these combinations will indicateif there are interaction effects.
MGMT 662: Integrative Research Project Analysis of Variance (ANOVA)
top related