Lecture 10- Repeated Variance Flashcards
What is systematic variance
Created by our manipulation
Unsystematic variance
Variance created by unknown factors
Positives of repeated measures designs
- Unsystematic variance is reduced
- Less participants are needed
- More sensitive to experimental effects
Approaches to repeated measures designs and the GLM
- Assume sphericity (estimate it, adjust the degrees of freedom)
- Fit a different kind of model (a multilevel growth model)
What is the assumption of sphericity
The differences between pairs of groups should have equal variance
How is assumption of sphericity estimated
- Greenhouse-Geisser estimate, Ê
- Huynh-Feldt estimate, Ē
What does it mean if E = 1
Sphericity is perfect
What does it mean if E < 1
Sphericity is violated to some degree
How do you correct the effect of sphericity
Multiply df by the estimates
Given that E quantifies the deviation from the perfect sphericity we correct the df by
the degree to which sphericity is violated
How does the df getting smaller affect sphericity
Makes it harder for the test to be significant
If u were to forget about sphericity you would have to
Routinely apply the G-G correlation
How do you specify repeated measures with afex::aov4() function
(Rm predictors|id variance)
With the afex::aov_4() function how do you build in interaction plot
afex_plot()
What package is lmer() from
lme4
When the repeated measures model is too complex to fit, the simplest version instead
Treats the predictor variables as fixed
what does j represent in the GLM
Different participants
b0j represents
b0 + u0j
u0j= variability across different participants
In post hoc tests how do you work out if something is significant
If the p value is smaller than the p critical value
What is true about sphericity
- It is automatically met when a variable has only two levels
- If it is not met then it is remedied by adjusting the df by the degree to which the data are not spherical
What is the column ges in afex::aov_4
- Generalised partial eta-squared
- Explains the percentage of the variance in the outcome
What assumption of the linear model is automatically violated in repeated measures designs
Independence of errors