Multiple Regression

high leverage point

predictor x has extreme values

influential point

has large effect on the regression models prediction

True or False influential points typically have high leverage

True

Does high leverage correlate with an influential point

No

leverage

associated with the jth observation and how far it is from the other n-1 observations.

influence measurement

cooks distance if Di > 1 than influential.

Small values of AIC, BIC, CP meaning

the model is relatively precise

When to use fixed X variables

• in designed experiments where the researchers have control of the values of X

• If we treat the observed values of X as the only ones that matter

When to use random X variables

• observational studies where the outcomes from a natural process

• Bayesian Regression

partial correlation coefficient

measures correlation between two variables after controlling for the others.

• use this when third variable may influence the relationship

• helps isolate the relationship between two variables.