Week 8 - Mixed Subject Design

Question 1

Benefits of Within Subject Measures

Answer 1

Give high level of power with low number of participants

Increases sensitivity of the test (Increase likelihood that the test will find an effect

Question 2

Elements of Within subject design

Answer 2

Treated as a two-way design

Has a participant factor and a focal X factor (Or multiple)

Question 3

Error in Mixed Design

Answer 3

Contain 2-way subject and interaction effect (Confounded in error - only 1 person per level of subject factor)
Between-Subject variance removed by being able to include the participant factor (effect of being a particular person is removed)

Question 4

Subject Factor

Answer 4

Mean of participant in one condition (Has main effect and interaction with focal IV)

Model inter-subject variation explicitly

Question 5

Two way interaction effect

Answer 5

Represent the different effect that the treatment X has on the participant (How people respond differently)
- Treat this as error

Allow model variation effect across person (Why people initially differ on items)

Use this interaction term as the error term (Make error smaller and test more powerful)

Question 6

Mixed Subject design

Answer 6

Used in most clinical trial studies (Time - Within, Group - Between)

Question 7

Sphericity Assumption

Answer 7

Different variance across variables, variance and co variance is the same

Variance between levels should be equal

Tested via Mauchley test (If non sig then assume the variances are equal)
- Go straight to Within subjects box

Overly sensitive to large sample size and type 1 error

Not sensitive enough with small sample size (Type 2 error - Variance is there but we can;t pick it up)

If significant use either the Wilks or Pillais in Multivariate test OR corrected rows under sphericity assumed

Question 8

Within VS Between

Answer 8

Within - Every person gets the same condition (Time)

Between - Every person is in a different level of the condition (Gender, Age, Treatment)

Question 9

Issues With Within Subject Design

Answer 9

Practice effects
Randomisation of presentation order
- Practice effect go into error
- does not exert systematic effect but will lower power

Counterbalancing - each participant get one of the predetermined sets of sequences that adjust presentation order to avoid practice and carry-over effects (Can calculate this and extract from error)

Question 10

Compound Symmetry Assumption

Answer 10

Strong assumption for within designs
Need both homogeneity of variance and correlation across levels of X
- Variance - Equal SD across levels of focal X are identical
- Correlation - Equal correlation between pairs of scores
Easily violated (When everyone agree on easy tasks)