LECTURE 6: ANOVA Flashcards

1
Q

ANOVA tests the null hypothesis that

A

means are the same
meanA=meanB=meanC (etc)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is an omnibus test?

A

overall test-test of number of different comparisons at once
*will say there is an overall difference! tells that means are diff, but not exactly which means are diff

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

one way independent ANOVA

A

3 independent groups, 1 DV
ex.
1. placebo
2. viagra
3. high viagra
DV: libido

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

what does F(2, 12) = 5.12 mean?

A

F =ANOVA statistic
2 = 2 groups
12 = sample n
5.12= calculated F

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

planned contrast is…

A

a priori test to determine where difference is!
statistical advantage, but can’t change later. ex. compare placebo to group1 and placebo to group3

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

how can you find where group differences lie in ANOVA?

A
  1. planned contrasts (1 group to 2 and 1 to 3)
  2. post hoc comparisons (not planned, compares all pairs of means!)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

post hoc comparisons are…

A

pairwise comparisons of all
designed to control alpha level (overall error rate)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Bonferroni and Tukey control for

A

type 1 error
*used if only small deviations from normality, sample sizes are equal, HOV met
(B more power than T if less pairs, Tukey more power than B if more pairs)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

If ANOVA is significant and variances between pairs are not homogeneous, use ____ post-hoc comparison

A

Games-Howell
accurate with unequal sample sizes (but better with larger samples)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

If data doesn’t meet assumptions in independent one way ANOVA, use ____

A

Kruskal-Wallis test

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

If there is more than 1 independent variable, it is called a

A

FACTORIAL ANOVA

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

If parametric assumptions are met, then independent one-way ANOVA. If significance found, do

A

post-hoc comparisons: bonferoni or tukey (not HOV: games-howell)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

If independent one-way ANOVA does not meet assumptions, do

A

Kruskal-Wallis
post hoc comparison: Mann-Whitney U (nonparametric equivalent to independent t-test)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

If assessing 1 group over time (3 or more times), must meet __ assumption

A

SPHERICITY: variances of differences btwn all possible pairs of groups are equal.
then do Repeated measures ANOVA
Bonferroni correction

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

what is sphericity?

A

variances of differences between all possible pairs of groups are equal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

what test do you use for sphericity?

A

Mauchley’s test
*assumption for RM one-way ANOVA

17
Q

If RM one way ANOVA assumptions not met, what is non-parametric equivalient?

A

Friedman’s Anova
post-hoc comparisons: Wilcoxon Signed Ranks

18
Q

what is the benefit of factorial ANOVAS?

A

can look at how variables interact
(interactions look for relationships btwn independent variables)
*effeect of IV on another IV

19
Q

The 3 tests for an omnibus factorial ANOVA

A
  1. interaction
  2. main effect for 1 IV
  3. main effect for 2nd IV
20
Q

If interaction is significant in independent ANOVA, then run

A

simple effects (forget main effects)
*adjust alpha level!

21
Q

If main effects are significant but interaction is NOT, then run

A

follow up main effect post-hoc comparisons and ADJUST ALPHA (3 groups = run bonferoni
OR NO FOLLOW UP IF ONLY 2 LEVELS

22
Q

what is a covariate?

A

extra/confounding variable that can affect outcome variable
(will strengthen analysis and more likely for you to find significance if you account for it: ANCOVA)

23
Q

when do you use MONOVA?

A

when testing for differences btwn groups when you have several dependent variables
CONTROLS TYPE 1 ERROR (takes into account relationships btwn dependent variables)

24
Q

what is the next step after significant global manova?

A

discriminant function analysis
(determine with of the DVs are best at separating the independent variables)