Between subjects design Flashcards
What is a between subjects design?
The research design where you assign a subject to a condition. Meaning that there is one data point for each variable
How many times you measure that subject here?
Once. No replications here needed because since you assign a subject to each condition.
What does replication mean?
That you have multiple subjects assigned to each one of your conditions. You can’t estimate the residual variance in this case
How does the simplest design would look like?
One categorical IV or predictor. One continuous IV or predictor and your outcome variable
What is the difference between an IV and a predictor
The experimental manipulation
If participants were not randomly assigned to different groups…
Is not an experimental design and you can create an imbalance. Number of people per condition unevenly distributed. Multicollinearity issues. Confounds added
In an F distribution you
Have the mean squares between divided by mean squares within.
Error variance
The variables that are uncontrolled for your study. Random noise
Nuisance variables
Another term to define variables that are just uncontrolled. people different ages, brain state. They create static in the signal that you are trying to detect.
Confound variables vs. Nuisance
Confound is anything varying along with our independent variable. You can’t tease them apart because they go along with each other. Like awareness and time of the day you assess
Notation
With multiple categorical predictors
m x n
or
m x n x o x p
Fractional factorial design
Not all conditions are run for resources and efficiency. Can test for main effects and low level interactions but not high level interactions
Problem with aggregation
The model assumes is error free (not true because you are increacing uncertainty when doing means or whatever you do)
Artificial increase or R square
No information about number of values that were averaged
Ecological fallacy
A probelm from aggregation where the average may not be representing each subject or trial
