You are on page 1of 2

Regression Diagnostic

Problem
Heteroscedasticit
y

Outliers

Leverage

What it
means
Error variances
are not equal

Extreme
values of Y
variable
Extreme
values of X
variable

Impact
Possibly
biased
standard
errors
(under or
over
estimate)
Poor fit

Slope
impact

Influence

Typically
extreme in X
or Y or both

Poor fit,
Slope of line
is changed

Multicollinearity

Independent
variables
exhibiting high
correlation

High R^2 ,
but many
independen
t variables
not
significant

Indicator/
Visual
Residual vs
fitted values
plot check for
patterns
absence of
pattern is good
Low R^2 ,
High residuals
Leverage vs
residual plotdiagonal value
of hat matrix
thumb rule
should not be
more than
2*p/n
Cooks distance
it combines
both outliers
and leverage thumb rule
should not be
more than 4/
(n-p-1)
Scatter plot
and Vif vif
should not be
more than 5.

p is number of parameters plus the constant term to be estimated


n is the number of observations

You might also like