Beast of Bias Flashcards
what are the four assumptions of the linear model?
linearity and additivity, sphericity, homoscedasticity, normality of residuals
which assumptions are the most and least important?
most important: linearity
least important: normality
what is linearity and additivity?
the relationship between the outcome and the predictor should be linear
what is sphericity?
error should be independent
what is homoscedasticity?
equal variances across predictors
what is normality of residuals?
the residuals should be normally distributed
how do you test for linearity?
plot of observed vs predicted should be linear
how do you test for sphericity?
plot of ZRESID vs ZPRED / any number >3 or
how do you test for homoscedasticity?
graph should not be funnel shaped or curved / the Robust F Statistic, welch and Brown Forsythe should share the same conclusion / variance ratio, okay if number is <2 / don’t use Levenes
how do you test for normality?
PP and QQ plots, case wise diagnostic, cook’s distance
what do you do for a case wise diagnostic?
standardised residuals which exceed 2 or 2.5 divided by the total number of cases, multiplied by 100
95% should lie between 2
99% should lie between 2.5
what do you do for cook’s distance?
any cases >1 are outliers / influencing the mean
what is bootstrapping?
constructs a CI based on the data by sampling the amount of scores in the data set with replacement
what are outliers?
atypical cases at odds with the data / impact the mean
