Lecture 6 Flashcards
Which test compares 2 means and only has 1 IV?
t-test
Which test compares 3+ means and can be used when you have manipulated more than one IV?
ANOVA or “factorial!”
When I use a lot of t-tests to compare several means, this inflates the ______
alpha
OR
Type 1 Error Rate
What is another name for ANOVA test?
Omnibus test
What does ANOVA do?
- tests for overall difference
- tells us that means are different
- doesn’t tell us exactly which means differ
What is an “overall test” that tests a number of different comparisons at once?
Omnibus
What question does ANOVA test?
Is there a difference somewhere?
What is the SS?
sum of squares or total amount of variance
What is the MS equation?
MS = Mean Square
MS = SS/df
F = ?
MS (between) / MS (within)
Why do we use follow-up tests?
Omnibus tells us that there IS a difference among the means, not WHERE
Follow up tests tell us WHERE the group differences lie
What is a planned contrast (follow up) test?
- Has statistical advantage
- Type of contrasts are limited
- Needs to be planned a priori
What is a post hoc comparison (follow up) test?
- MOST common
- not planned
- compares ALL pairs of means
What is a pairwise test?
Typically after global test & implies ONLY 2 means
What is simple effects?
run this after global test & significant interaction
Which tests can we use if there are only (1) small deviations from normality, (2) sample sizes in each cell are equal, and (3) variances are homogenous?
Bonferroni & Tukey
Which tests are conservative and controls for type 1 error?
Bonferroni & Tukey
Bonferroni has more power than Tukey if ______
number of comparisons are small
Tukey has more power than Bonferroni if _____
the number of comparisons are large
and more power than Dunn and Scheffe
Which test do you use if variances are not homogenous?
Games-Howell
What is the Games-Howell?
Powerful but can be liberal when sample sizes are small. Accurate with unequal sample sizes
What is the Kruskal-Wallis Test?
The non-parametric equivalent to an independent one-way ANOVA
- select if your data doesn’t meet assumptions
- Mann-Whitney U tests
What is the HOV assumption equivalent in ANOVA called?
Sphericity
What is the sphericity assumption?
The variances of the differences between all possible pairs of groups are equal
What is the test for sphericity called?
Mauchley’s test
When you have F(2,10) = 12.5,
what does the 1st number mean?
what does the 2nd number mean?
1st number: # of groups
2nd number: # of participants per group
What is Friedman’s ANOVA?
non-parametric equivalent of RM - one-way ANOVA
What is the difference between a two-way independent ANOVA vs. a factorial ANOVA analysis?
2-way independent ANOVA has 2 IV’s
factorial ANOVA has 1 IV
What does “interaction” look for?
Interesting relationships between the independent variables
T/F: If the interaction is significant, don’t worry about the main effects.
True
If the interaction lines are parallel, is it significant?
No
When the interaction is significant, you can run ___ ____
simple effects
If main effect(s) is significant and interaction is NOT significant, you should run ________
Follow-up main effect pairwise comparisons & report. Adjust alpha level!
Note: if only 2 levels for main effect, no follow-up testing needed
If main effect(s) is significant AND interaction is significant, you should run _____
Simple effects and “forget about” main effects. Adjust alpha level!
When do you run simple effects?
If you have a significant interaction!
If I am studying self-perceived preparedness in Dental students. I have 3 independent variables and 1 dependent variable. What type of test should I run?
Split/Mixed 3-way Factorial ANOVA
If I am studying self-perceived preparedness in Dental students. I have 3 independent variables and 1 dependent variable.
How many main effects? How many 2-way interactions? How many 3-way interactions?
3 Main Effects
3 2-way Interactions
1 3-way Interaction
What is a covariate?
A variable that is complementary or related to the dependent variable; extraneous variable
When would you test for covariates?
Test differences between group means when we know that an extraneous/confounding variable can affect the outcome variable
WHY would you test for covariates?
To strengthen your analysis
More likely to find significance if you can identify the covariate and measure it beforehand
What is ANCOVA?
analysis of covariance
run analysis on the adjusted means
What is MANOVA?
Multivariate analysis of variance
When do you use MANOVA?
To test for differences between groups when we have several DVs
Why is MANOVA better than multiple ANOVAs?
- controls familywise error rate (Type 1 error)
Note: takes into account relationship between the DVs; more powerful than separate ANOVAs
DVs are mathematically combined into aggregate turning it into one variable for the global
What’s the next step after a significant Global MANOVA?
DFA (Discriminant function analysis)
What is DFA (Discriminant function analysis) used for?
Used to determine which of the DVs are best at discriminating the groups
