Chapter 12 psych stats

Analysis of variance (ANOVA) is a form of hypothesis testing that evaluates the mean
differences between two or more treatments. The main difference between t tests and anova
tests is that t tests can only test up to two treatments while anovas can test more than two
treatments. Anova can be used with independent measures and repeated measures designs. And,
anovas can be used when there is more than one independent variable.
Independent variables – a variable the researcher manipulates to create differences
Factors- independent variables in analysis of varaince testing.
Single factor design- one factor (independent variable) in the study
Factorial design- more than one factor (independent variables) in the study
Single factor, Independent measures design
The goal of the researcher is to decide if
a. there are no differences between populations and the samples are different due to
chance.
b. there are real differences between the populations
Hypotheses for ANOVA
Null: no treatment effect, IV had no effect, no differences between populations
Alternative: there is a treatment effect, IV had an effect, at least one population is different from
the other populations ( this alternative hypothesis is not specific)
Test Statistic: F-ratio
The F-ratio is the test statistic used for ANOVA, it is based off of variance instead of mean
differences because the anova sometimes looks at situations where there are more than two
sample means.
The variance (numerator) provides a single number that represents the differences between all
sample means. The variance in the denominator shows the variance that occurs due to chance
much like the standard error in t statistics, this number in the F-ratio is called error variance.
Logic of ANOVA
To measure the amount of variability and where it comes from
1. Between Treatment Variance: variability between treatments in order to provide a
measure of the overall differences between treatment conditions (difference between
sample means
2. Within Treatment Variance: variability inside each treatment condition
Between treatment variance:
a. variance due to real differences
b. variance due to chance

Within treatment variance:
a. variance due to chance
The F-ratio : compares between and within variance
If there is NO effect both the numerator and denominator will measure chance, so they should be
close to equal so the F-ratio will be close to 1.00.
If there IS an effect the numerator should be larger than chance (denominator) and the F-ratio
should be larger than 1.00.
Notation:
Levels- the individual treatment conditions that make up a factor (different values used for the
IV)
k- number of treatment conditions (levels) of a factor
n- number of scores in each treatment
N- total number of scores in the entire experiment; when the samples are the same size N=kn
T- sum of X for the each treatment condition
G- grand total, sum of all scores in the experiment, by adding up the treatment T’s
Sum of Squares
1. Total Sum of Squares, SStotal- the sum of squares for the entire set of N scores
2. Within-Treatments Sum of Squares, SSwithin- variability within the treatments; add all the
individual values of SS for each treatment condition
3. Between-Treatment Sum of Squares, SSbetween- variability between treatments
Degrees of Freedom
1. Total Degrees of Freedom
2. Within-Treatment Degrees of Freedom
3. Between-Treatment Degrees of Freedom
Variance:
Mean square (MS)- is the term used for variance in ANOVA, since the variance is the
mean squared deviation
Distribution of F-ratios:
When the null is true the numerator and denominator are measuring the same thing (chance
variance) because the treatment did not have an effect and the F-ratio is around 1.00.
a. F values are always POSITIVE because they are computed from variances which
are positive. (cut off at zero)
b. F-values should pile up around 1.00, when the null is true.

With large df, the F values will cluster around 1.00, with small df, the values are more spread
out.
Table: Appendix B
Degrees of freedom for numerator along the top
Degrees of freedom for denominator along the side
Values in the table are the critical regions at the .05 (top row) and .01 level (bottom row)
Post hoc tests:
When you reject the null you conclude the means are not all the same, must decide which ones
are different, so you compare two treatments at a time called: pairwise comparisons.