R Squared And Violations Of Clrm Flashcards
R squared
Goodness of fit- how well the regression fits the data points
Adjusted r squared
Takes info accounting a higher r squared if more variables have been added
So adjusted r squared falls if more variables are added
Violation- errors have mean zero
Only true if there’s a constant teen in regression so it’s fits data better
Violation- variance in errors is constant
Homescadacity isn’t true if there’s patterns in the errors
Test for heteroscedacity
Whites test- checks for a pattern between errors
Violation- errors normally distributed
Test coefficient skewness- berd jarque
To see if curve is skewed
Violation- correlation between errors
Multicolinerity
E.g if x3= 2x2
R squared will be high
Solution- collect more data
Don’t remove important variables- bias
Violation- no pattern between errors
Test for autocorrelation Detect- durbin Watson test- assumes errors are autocorrelated
Patterns in residuals could be positive - curve or negative- straight jumps through x