One-factor ANOVA Practical
Module Overview
Welcome to Module three of the Analysis of Variance series.
Focus: One factor analysis of variance and conducting these analyses in SPSS.
Assumptions of Basic Analysis of Variance
Introduction to the four key assumptions:
Dependent Variable Scale:
Must be of an interval or ratio scale.
Normal Distribution:
Dependent variable scores should ideally be normally distributed.
Can be slightly violated if sample sizes exceed 30.
Equivalent Group Sizes:
Group sizes should be roughly equivalent, with a tolerable ratio of about 1.5.
Homogeneity of Variance:
Assumes the variance among populations is equal across group means, though sample variances may differ slightly.
Detailed Exploration of Homogeneity of Variance
Variance Definition:
Measures the spread of data relative to the mean.
Variance Example Calculation:
For sample variance:
Formula: ext{Variance (s^2)} = \frac{\sum (X - \text{mean})^2}{N - 1}
Example with data: 2, 3, 4
Mean = 3; Variance = 1.
Example with data: 0, 3, 6
Mean remains 3; Variance = 9.
Significance of Variance:
Variance calculations should be similar for all groups in ANOVA to meet the homogeneity assumption.
Implications of Violating Assumptions:
If homogeneity is violated, it necessitates conducting nonparametric tests as a last resort.
Analysis by Hand - One Way ANOVA
Review of calculating sum of squares by hand using:
Data from Module Two (as a reference).
Steps for calculating:
Calculate totals for each group.
Calculate sum of squares for each group and total sum of squares.
Calculate the degrees of freedom.
Example calculations:
Total participants: 50
Total sum of squares result: 768.82.
Sum of squares between: 351.52 with 4 degrees of freedom.
Sum of squares within: Calculate as total - between.
Analysis of Variance Table
Structure includes:
Rows for between and within groups, total.
Calculation of mean square: \text{Mean Square} = \frac{\text{Sum of Squares}}{\text{Degrees of Freedom}}
F-value comparisons against critical values to test significance (with example values).
Result interpretation:
Example: F-value = 9.1 exceeds critical value = 2.61, thus null hypothesis rejected.
SPSS Analysis of Variance
Steps to initiate analysis in SPSS:
Input data under States of independent variable with levels (low, medium, high).
Steps to enter into SPSS for analysis of variance:
Navigate to ANALYZE > General Linear Model > Univariate.
Select factor and dependent variable.
Enable options for descriptive statistics and homogeneity tests.
Output Interpretation from SPSS
Identification of descriptive statistics,
Levene's test results, indicating normality and significance.
ANOVA table including:
Significance of F = 112, p < 0.001.
Reporting format: F(2, 15) = 112.82, p < 0.001.
Reporting partial eta squared (93.8% variance explained).
Discussing observed power of the analysis.
Post Hoc Tests
Explanation of Bonferroni test to avoid family-wise error when performing multiple comparisons.
Results show significant differences across all tested conditions with relevant p-values.
Discuss the relevance of graphical representation for visualizing means among groups.
Concept of Family-Wise Error
Definition: Probability of making at least one type I error across multiple tests.
Explanation of how repeated t-tests without correction may lead to inflated error rates.
Summary of the Bonferroni correction process to maintain family-wise error rate.
Planned Comparisons and Contrast
Explanation of planned comparisons as specific comparisons made prior to analysis to control for family-wise error.
Types of contrasts differentiating orthogonal (independent) vs. non-orthogonal (dependent) comparisons.
Trend Analysis
Purpose: Exploring systematic changes in dependent variable across levels of independent variable.
Examples using temperature and anger levels to identify linear vs. quadratic trends.
Within Groups vs. Between Groups ANOVA
Review of within-subjects design focusing on participants experiencing all levels of IV.
Discuss covariates and sphericity adjustments in the composition of within ANOVA table outputs.
Details on Mauchly's test and corrections applicable if sphericity is violated.
Conclusion
Summary of key takeaways from one factor ANOVA, application in SPSS and introduction into future topics relevant to ANCOVA, analysis of covariance.