7118 Lecture 3: Lecture 3B
20 Jun 2018
School
Department
Course
Professor
LECTURE 3B
One way repeated measures ANOVA
Basic concepts
- Also called related anova, correlated scores anova, related samples anova, matched analysis
anova
- One way means there is only one IV
- Use this test when there are 3+ variables of single IV and it is manipulated within subjects ie
same people in different conditions
- Repeated measures version of the one way between groups
- Both involve IV with several groups and one DV.
- Used when tracking people over time
oIe before, midway, after, follow up
- Someties we are interested in diff conditions eg driving ability of same people on wet vs dry vs
foggy vs dirt road
- Can also use it when there are different people in different conditions, but they are matched in
some way eg husband and wife, or matched on IQ
- Don’t use repeated t tests because of the same problems with error for the btw groups anova
instead we test too see if there is a difference anywhere btw groups of scores
- Same number of people in each group – bc same people
Why use repeated measures?
- Statistical reasons: easier to find a significant difference – by collecting data from the same
participants under repeated conditions the individual differences can be eliminated or reduced
as a source of between group differences
- Economic/logistical reasons: this design also proves to be economical in terms of amount of
time/money spent to recruit people, especially when sample members are difficult to recruit
Still only one dv
- Need only one dv to conduct it
- Can only be used when you are measuring the same characteristic under different conditions
- Can manipulate something between different conditions as long as you are measuring the same
characteristic eg racism against aboriginal, greek, indian etc
What samples to use a repeated measures anova
- Timepoints
oBefore, mid, after semester
- Environments
oHigh, medium, low, now
Yes for same people
No for different people – one way anova
- Age groups
oNope; unless follow them throughout their life
- Treatments
oCognitive behavioural therapy etc
oNo; too much likelihood of them affecting people in the wrong way
Same as one way between anova
- Start with null and alternative hypothesis
- Calculate F by dividing MS treatment by MS error
- Two df
- Find p value
- If p is smaller than alpha, reject null hypothesis
- Can calculate effect size, type 1 and 2 error and power
- Need extra step at end to find out where the difference is
Whats new
- Calculate an extra bit of variance – the btw subjects variance (bit that is the same about a
person over repeated measurements)
- Take it out of the MS error so that the MS error is smaller
- Extra step at end is different; but simpler
Individual differences
- Indidivual differences tend to be constant over time; we get rid of this
- In the btw groups anova, we didn’t know that bit that is special to the person
Structural model
-
- Eg happiness rating in three diff place; home, uni, café
oAverage rate is 10
oYours is 11
oAverage effect for home is +2, uni is -4, café is +5
-
-
-
-
One way repeated measures anova
- Iv has more than two groups and has been manipulated within subjects or have matched
samples
Example
Document Summary
Also called related anova, correlated scores anova, related samples anova, matched analysis anova. One way means there is only one iv. Use this test when there are 3+ variables of single iv and it is manipulated within subjects ie same people in different conditions. Repeated measures version of the one way between groups. Both involve iv with several groups and one dv. Used when tracking people over time: ie before, midway, after, follow up. Someties we are interested in diff conditions eg driving ability of same people on wet vs dry vs foggy vs dirt road. Can also use it when there are different people in different conditions, but they are matched in some way eg husband and wife, or matched on iq. Don"t use repeated t tests because of the same problems with error for the btw groups anova instead we test too see if there is a difference anywhere btw groups of scores.
