One-Way Analysis of Variance

⚫ The purpose of a one-way analysis of variance

(one-way ANOVA) is to compare the means of two or
more groups (the independent variable) on one
dependent variable to see if the group means are
significantly different from each other
When to Use One-Way ANOVA?
⚫ There is one independent variable (factor) with two
or more groups (e.g., "type of diet" with three groups:
low-carb, low-fat, and keto).
⚫ There is one dependent variable that is continuous
(e.g., "weight loss in pounds")
why we use one-way ANOVA?
⚫ the independent t-test and one-way ANOVA are
similar because they both compare group means,
there’s a key reason why we use one-way ANOVA
instead of multiple t-tests when comparing three or
more groups:
⚫ Multiple Comparisons Increase the Risk of Type I
Error
⚫ Every time you perform a t-test, there is a chance
(usually 5%, or α=0.05) of making a Type I error
(rejecting the null hypothesis when it is true).
⚫ If you conduct multiple t-tests, the probability of
making a Type I error increases cumulatively with
each test.
⚫ For example, if you compare 3 groups, you would need
to conduct 3 t-tests. For 4 groups, you’d need 6 t-tests.
⚫ Each additional t-test increases the overall error rate,
making your results unreliable. This is known as the
family wise error rate (FWER).
⚫ One-way ANOVA solves this issue by testing all
group means simultaneously in a single analysis,
keeping the Type I error rate controlled at the desired
level (e.g., 0.05).
⚫ Efficiency and Simplicity
⚫ ANOVA is more efficient because it analyzes all group
differences at once rather than requiring separate
calculations for each pair of groups.
⚫ Interpretation and Post Hoc Tests
⚫ One-way ANOVA gives a global test of significance (it
tells you if there’s a significant difference somewhere
among the group means).
⚫ If the ANOVA result is significant, you can then run post
hoc tests (e.g., Tukey’s test or Bonferroni correction) to
figure out which specific groups differ.
Assumptions for ANOVA
⚫ Assumption of Normality
⚫ The dependent variable should be normally distributed
within each group of the independent variable.
⚫ This assumption ensures that the statistical tests used in
ANOVA are valid.
⚫ Homogeneity of Variance
⚫ Also called the assumption of equal variances, this
means the variance (spread) of the dependent variable
should be roughly equal across all groups.
⚫ This ensures that the group comparisons are fair and
unbiased.
⚫ Independence of Cases
⚫ The observations in each group must be independent of
each other (i.e., no overlap or repeated measures).
⚫ This ensures that there is no bias or duplication in the
data.
⚫ Why Are These Assumptions Important?
⚫ Violating these assumptions can lead to:
⚫ Invalid results (e.g., incorrect p-values).
⚫ Increased Type I or Type II errors (false positives or
false negatives).
⚫ Misleading interpretations of the data.
One-Way ANOVA in Depth
⚫ Purpose of One-Way ANOVA
⚫ The primary goal of a one-way ANOVA is to determine
whether the means of two or more groups differ
significantly by dividing the total variance in the
dependent variable into two components:
⚫ Between-Group Variance: The variation due to
differences between the group means.
⚫ Within-Group Variance (Error): The variation due
to differences within each group, often referred to as
random error.
How It Works
⚫ When analyzing data:
⚫ Mean as the Best Estimate:
⚫ A sample mean is our best guess of the population mean.
⚫ Any deviation of individual scores from the sample mean
is attributed to random error.
⚫ Random Error:
⚫ When we randomly select individuals from a population,
we expect variability in scores.
⚫ This variability within groups is called within-group
error or mean square within (MSw).
⚫ The critical question here is:
⚫ Does the between-group variance (differences between
group means) exceed the within-group variance
(random error)?
⚫ If between-group variance is significantly larger than
within-group variance, it suggests that the group means
are different.
⚫ Statistical Significance:
⚫ Compare the F value to a critical value in the F
distribution table (based on degrees of freedom).
⚫ If the F value is large enough, the differences between
group means are considered statistically significant.
Example
⚫ Imagine we test basic math skills for three groups of 5th
graders.
⚫ Group A: Average score = 12
⚫ Group B: Average score = 15
⚫ Group C: Average score = 18
⚫ Between-Group Variance: Differences between the group
averages (12, 15, 18).
⚫ Within-Group Variance: Differences among scores
within each group.
⚫ If the between-group variance (MSb) is much larger than
the within-group variance (MSw), ANOVA will detect
significant differences between the groups.
⚫ One-Way ANOVA vs. Independent t-Test
⚫ Similarity to the t-Test:
⚫ The numerator of the t-test formula represents the
difference between two sample means. In ANOVA, this
concept expands to measure the average differences
between multiple group means (Mean Square Between,
MSb).

⚫ The denominator of the t-test is the standard error of the

difference, which is equivalent to the average
within-group error in ANOVA (Mean Square Error,
MSe).
⚫ Key Differences:
⚫ Number of Groups: A t-test compares two means, while
ANOVA is designed to handle two or more means.
⚫ Squaring the Differences: ANOVA uses squared
differences to calculate the sum of squares for
within-group (SSe) and between-group (SSb) variability.
SSe
SSb
Key Components in ANOVA
⚫ Sum of Squares (SS):
⚫ ANOVA divides variance into two parts:
⚫ SSb (Between-Groups): Measures how much group means differ from
the overall mean (grand mean).
⚫ SSe (Within-Groups): Measures variability of individual scores within
each group.
⚫ Mean Squares (MS):
⚫ MSb (Mean Square Between): The average squared
deviation between group means.
⚫ MSe (Mean Square Error): The average squared deviation
within groups.
⚫ Sum of Squares Total(SSt)
⚫ Sum of the squared deviations between individual scores and the grand
mean on the dependent variable
⚫ Sum of Squares Between + Sum of Squares Error
Example of mean score
ETA Squared
⚫ A measure of the association
between the independent
variable and the dependent
variable

⚫ SSB : Sum of Squares Between

Groups
⚫ SST : Sum of Squares Total
Post Hoc Tests
⚫ Post Hoc Tests: Understanding the Basics
⚫ After finding a statistically significant F value in
ANOVA, our work isn’t finished. While the F value
tells us there’s a difference between group means, it
doesn’t tell us which specific groups differ. This is
where post hoc tests come in.
⚫ Why Do We Need Post Hoc Tests?
⚫ When comparing multiple groups (e.g., 3 or more), a
large F value indicates a significant difference
somewhere, but it doesn't show where.
⚫ Post hoc tests compare each group mean to every
other group mean to pinpoint the significant
differences.
⚫ How Do Post Hoc Tests Work?
⚫ Post hoc tests calculate:
⚫ The difference between two group means.
⚫ Whether that difference is statistically significant, by
dividing it by a standard error (a measure of
variability).
⚫ Different post hoc tests use different formulas for
calculating the standard error, which is why their
results can vary.
⚫ Example of a Post Hoc Test: Tukey HSD
⚫ Tukey HSD (Honestly Significant Difference) is one
of the most common post hoc tests.
⚫ It’s considered a liberal test, meaning it’s more likely
to detect significant differences between group means
compared to conservative tests like the Scheffe.
⚫ Tukey HSD assumes that all groups have the same
number of cases (sample size).
Steps for Using Tukey HSD
⚫ Calculate the observed Tukey HSD value for each pair of
group means.
⚫ This involves subtracting one group mean from another and
dividing by the standard error.
⚫ Compare the observed value with a critical value:
⚫ The critical value depends on the number of groups and the
degrees of freedom for error (df error).
⚫ These critical values are usually found in a table (similar to an
F table).
⚫ Interpret Results:
⚫ If the observed Tukey HSD value is larger than the critical
value, the difference between the two groups is statistically
significant.
⚫ When Should You Use Tukey HSD?
⚫ Best for equal sample sizes: Tukey HSD assumes all
groups are the same size.
⚫ If groups have unequal sizes, other tests like the
Scheffe or Games-Howell may be better.
SPSS Activity
⚫ Analyze> Compare Means> One Way ANOVA
Write up
I performed a one-way ANOVA to compare the three
groups’ average liking of bubble gum ice cream. This
analysis produced a statistically significant result
(F(2,12) = 4.94, p < .05).
(If you run Post hoc )
Post hoc Tukey tests revealed that the only significant
difference between groups was
found between 5-year-olds (M = 4.20) and 12-year-olds
(M = 2.40), with the younger children
liking bubble gum ice cream significantly more than the
older children.”
paired t test
⚫ A paired t test was calculated to compare the grade
point average (GPAs) of students when they were in
5th grade and a year later when they were in 6th grade.
The analysis produced a significant t value (t(688) =
8.19, p < .001). An examination of the means revealed
that students had higher GPAs in 5th grade (M = 8.08)
than they did in 6th grade (M = 7.35).”
Independent sample t test
⚫ I performed an independent t test to compare the grade
point averages (GPAs) of 6th grade boys and girls. The
analysis produced a significant t value (t(708) = –7.45,
p < .001). An examination of the means revealed that
boys had lower GPAs (M = 6.58) than did girls (M =
8.14).
correlation
⚫ The correlation between GPA and test scores was
positive, moderately strong, and statistically significant
(r =.43, p < .001). The coefficient of determination (r
sqaure = .18) revealed that 18% of the variance in test
scores was explained by students’ GPA.”
multiple regression analysis
⚫ A multiple regression analysis was conducted to
examine the predictors of self-handicapping.
⚫ Four predictors were simultaneously entered into the
model: Avoidance Goals, GPA, Gender, and
Self-efficacy. Together, these predictors accounted for
11% of the variance in self-handicapping.
⚫ All of these variables except for gender were
significant predictors of self-handicapping. GPA (β =
–.21) and self-efficacy (β = –.20) were the strongest
predictors and were negatively associated with
self-handicapping, whereas avoidance goals were
positively associated with self-handicapping (β = .13).