One-Way and Factorial ANOVA - PowerPoint PPT Presentation

About This Presentation
Title:

One-Way and Factorial ANOVA

Description:

Method #1: Compare Means. First we have to test if we meet the assumptions of ANOVA: ... Method #1: Compare Means. One-Way ANOVA. Analyze Compare Means One-Way ... – PowerPoint PPT presentation

Number of Views:124
Avg rating:3.0/5.0
Slides: 27
Provided by: michae46
Category:
Tags: anova | factorial | method | one | way

less

Transcript and Presenter's Notes

Title: One-Way and Factorial ANOVA


1
One-Way and Factorial ANOVA
  • SPSS Lab 3

2
One-Way ANOVA
  • Two ways to run a one-way ANOVA
  • Analyze ? Compare Means ? One-Way ANOVA
  • Use if you have multiple DVs, but only one IV
  • Analyze ? General Linear Model ? Univariate
  • Use if you have only one DV bc/ can provide
    effect size statistics
  • More on this later (factorial ANOVA section)

3
Method 1 Compare Means
  • First we have to test if we meet the assumptions
    of ANOVA
  • Independence of Observations
  • Cannot be tested statistically, is determined by
    research methodology only
  • Normally Distributed Data
  • Shapiro-Wilks W statistic, if significant,
    indicates significant non-normality in data
  • Analyze ? Descriptive Statistics ? Explore
  • Click on Plots, make sure Normality Plots
    w/Tests is checked

4
Testing Assumptions
5
Testing Assumptions
  • Homogeneity of Variances (Homoscedasticity)
  • Tested at the same time you test ANOVA
  • Analyze ? Compare Means ? One-Way ANOVA
  • Click on Options and make sure Homogeneity of
    variance test is checked
  • If violated, use Brown-Forsythe or Welch
    statistics, which do not assume homoscedasticity

6
Method 1 Compare Means
  • One-Way ANOVA
  • Analyze ? Compare Means ? One-Way ANOVA
  • Dependent List DVs Factor IV
  • Options
  • Descriptive
  • Fixed and random effects
  • Homogeneity of variance test
  • Levenes Test Significant result ?
    Non-homogenous variances
  • Brown-Forsythe
  • Welch
  • Means plot

7
Method 1 Compare Means
8
Method 1 Compare Means
9
Method 1 Compare Means
  • One-Way ANOVA
  • Post-Hoc
  • Can only be done if your IV has 3 levels
  • Pointless if only 2 levels, just look _at_ the means
  • Click the test you want, either with equal
    variances assumed or not assumed
  • DONT just click all of them and see which one
    gives what you want (thats cheating), select the
    test you want priori

10
Method 1 Compare Means
  • Contrasts
  • Click Polynomial, Leave Degree at default
    (Linear)
  • Enter in your coefficients
  • of coefficients should equal of levels of
    your IV
  • Doesnt count missing cells, so if you have 3
    levels, but no one in one of the levels, you
    should have 2 coefficients
  • Coefficients need to sum to 0

11
Method 1 Compare Means
  • Contrasts
  • Enter in your coefficients
  • IV Race 1Caucasian, 2African American,
    3Asian American, 4Hispanic, 5Native American,
    6Other, BUT there were no Native Americans in
    the sample
  • If you want to compare Caucasians to Other,
    coefficients 1, 0, 0, 0, -1
  • Caucasians vs. everyone else -1, .25, .25, .25,
    .25

12
Method 1 Compare Means
13
Method 2 Univariate
  • Univariate works for both one-way (1 IV) and
    factorial ANOVAs (2 IVs)
  • Allows for specification of both fixed and random
    factors (IVs)
  • Assumptions
  • Independence of Observations
  • Normally Distributed Data
  • Both same as one-way ANOVA

14
Factorial ANOVA
  • Assumptions
  • Homoscedasticity
  • Tested at the same time you test ANOVA
  • Click on Analyze ? General Linear Model ?
    Univariate
  • Click on Options and make sure Homogeneity
    tests is checked

15
Factorial ANOVA
  • Options
  • Estimated Marginal Means
  • Displays means, SDs, CIs for each level of
    each IV selected
  • If Compare main effects is checked, works as
    one-way ANOVA on each IV selected
  • Confidence interval adjustments allows you to
    correct for inflation of alpha using Bonferroni
    or Sidak method
  • Descriptive statistics
  • Estimates of effect size
  • Observed power
  • Pointless, adds nothing to interpretation of
    p-value and e.s.
  • Homogeneity tests
  • Levenes test

16
Factorial ANOVA
17
(No Transcript)
18
Factorial ANOVA
  • Save
  • Dont worry about this for now
  • Post Hoc
  • Select the IV for which you wish to compare all
    levels against all other levels (i.e. that you
    dont plan to do planned comparisons on)
  • Click on the right arrow button so the IV is in
    the box labeled Post Hoc Tests for
  • Check the post hoc tests you want done, either
    with equal variances assumed or not assumed
  • Click Continue

19
(No Transcript)
20
Factorial ANOVA
  • Plots
  • Horizontal Axis
  • What IV is on the x-axis
  • Separate Lines
  • Separate Plots

21
Factorial ANOVA
  • The following graph has the IV Race on the
    horizontal axis and separate lines by the IV
    Gender

22
Factorial ANOVA
  • Model
  • Allows you to
  • Denote which main effects and interactions you
    are interested in testing (default is to test ALL
    of them)
  • Specify which type of sum of squares to use
  • Usually you wont be tinkering with this

23
Factorial ANOVA
  • Contrasts
  • Tests all levels within one IV
  • Concern yourself with Simple only for now
  • Reference category What level all others are
    compared to (either first or last, with this
    referring to how they were numbered)
  • Can test specific levels within one IV with
    specific levels in another IV, but requires
    knowledge of syntax

24
(No Transcript)
25
Factorial ANOVA
26
Factorial ANOVA
  • Interpreting interactions
  • See graphs
Write a Comment
User Comments (0)
About PowerShow.com