Week 5 Flashcards
R2 equation
variance of DV-variance of residuals /. total variance x 100
R2 groups
.04 small
.09 medium
.25 large
adjusted R2
Gives an estimate of how variability would be explained if the model was derived from population not sample
shrinkage
large discrepancy between R2 and adjusted R2
regression model does not generalise well to population
f2
effect size
proportion of residual variance explained
assumptions of linear regression
outcome= continuous
predictor= continuous/dichotomous
predictors must have non-zero variance
linearity
independent variables and errors
normally distributed errors/residuals with a mean of 0
equal variance/homoscedasticity
checking linearity
residuals vs predicted= absence of clear pattern
homoscedasticity
for each value of the predictors, the variance of error term should be constant
checking normally distributed errors
scatterplots- residuals clustered around regression line
histogram of standardised residuals- bell shapes
P-P plots of regression standardised residuals- on line
smaller CI
more precise estimate of true population value
wide CI
more uncertainty about the true value
SD for categorical predictors
harder to interpret
multicollinearity
high intercorrelations between predictors
multicollinearity checking
tolerance > .10
VIF < 10
sr2
square semi partial correlation
proportion of variability in the outcome uniquely accounted for by that predictor
