Week 1 MRA Flashcards
When does it make sense to perform a linear regression
Hoge correlaties? JA
Zie correlatieshoogste significante correlatie beste voorspeller
what should you check to determine the null hypothesis of no relationship can be rejected in liniar regression
ANOVA F, df and p
how to check how much total variance is explained in liniar regression
zie R2
What predictor explains the most unique variance?
see coefficients table–>part–>highest number
Is there evidence of multicollinearity in the predictors
check
* alle VIF <10
* alle tollerance >0.1
JA?–>geen
how to check if there are outliers on X
maximum centred leverage value (zie residuals statistics) <3(k+1)/N
k=aantal predictoren
how check if there are outliers on Y
min en max std residuals tussen -3 en 3–> geen outliers
how to check if the outliers are influential?
max cook’s distance <1–>non influential
How do you check for linearity?
scatterplot–>fit line total–>zie r2 linear bovenaan de grafiek.
how to check for homoscedasticiteit
scatterplot—>horizontale lijn in het midden–>spreiden boven en onder hetzelfde?–>homoscedasticiteit.
how to check for normality? What if the terms for normality are not met?
pp plot–>punten ongeveer op 1 lijn?
robust if N>100
How do you check if adding another variable improves the model in hierarichical regression?
model summary–>F change sign. noteer F, df, df p.
how to check how much extra variance is explained by the new model in hierarchical regression?
model summary–>R square change.
When do you use MRA
- predicting
- multiple X
- one Y
- all variables are Interval
