Categorical Predictors Flashcards
How is the GLM used in experiments?
Manipulation of variables is used to induce a difference
What does the F statistic do?
Calculate how much variability is explained by the linear model when we fit the model and how much cannot be explained - error
If the model explains more variability than it can’t, what will be significant?
The F statistic
The predictors will overall significantly predict the outcome
Why are experiments good?
Because we can establish cause and effect - this is due to manipulation
if the model is significant, it is the manipulation which has influenced the outcome
What does dummy coding refer too?
Coding groups you want to compare
Count number of groups - 1. this is the amount of dummy codes that you need
choose one group as baseline that you compare groups too
What dummy codes do you apply to each group?
0 = control group 1 = experimental group
What does beta 0 represent?
The mean in the condition coded 0 - this is the intercept
What does beta 1 represent?
The difference between the means
What does the T test do?
Gives the same values when carry out a linear model - it is a specific case of the linear model
How do you dummy code multiple categories?
Baseline category coded as 0
Compare each group against the baseline, so each exp group will be 1
The B for each dummy variable will be the difference in means between each category and the baseline
What are the sum of squares for categorical predictors?
Total - difference between overall mean and fear
Residual - differences between each observed point and the group mean
Model - looking at predicted value vs overall mean
What is R the ratio between?
The experimental effect to the background ‘error’ - signal-noise ratio
What are robust tests of equality of means?
Correct for the amount of heteroscadicity
small amount = small correction
they will be more correct than uncorrected
welch and brown-forsythe
How does the linear model accommodate categories?
By using dummy coding
