assumptions Flashcards
linearity assumption
relationship between x and y is linear
normality assumption
errors are normally distributed around each predicted value
equal variance assumption
variance is constant across values of predictors and fitted values y^
independence of errors assumption
errors are not correlated with each other
how is the linearity assumption assessed?
scatterplots with loess lines for single variables
component-residual plots for multiple predictors
- crPlots()
how is the normality assumption assessed?
qqplots
histograms
how is the equal variance assumption assessed?
plot residual values against predicted values y^
- residualPlot()
how is the independence of errors assumption assessed?
evaluated based on study design
why do interaction terms result in multicollinearity
because they are made up of the product of main effects
how can you reduced multicollinearity which comes from model specification rather than the data
mean centring the variables
what does multicollinearity refer to
correlation between predictors
what increases when there are large correlations between predictors
standard error
