Model Browser User's Guide

Definitions

Symbol	Definition
N	Number of data points
p	Number of terms currently included in the model
q	Total number of possible model parameters (q=p+r)
r	Number of terms not currently included from the model
y	(Nx1) response vector
X	Regression matrix. X has dimensions (Nxq)
Xp	(Nxp) model matrix corresponding to terms currently excluded from the model
Xr	(Nxr) matrix corresponding to terms currently excluded from the model
	(px1) vector of model coefficients
PEV	Prediction Error Variance
	User-defined threshold criteria for automatically rejecting terms
	(Nx1) vector of predicted responses.
e	(Nx1) residual vector.
e(i)	(Nx1) vector of PRESS residuals.
H	Hat matrix.
L	(Nx1) vector of leverage values.
VIF	Variance Inflation Factors
SSE	Error Sum of Squares. SSE = e'e
SSR	Regression Sum of Squares. SSE =
SST	Total Sum of Squares. SST = y'y - N
MSE	Mean Square Error. MSE = SSE/(N-p)
MSR	Mean Square of Regression. MSR = SSR/P
F	F-statistic. F = MSR/MSE
MSE(i)	MSE calculated with ith point removed from the data set.
RMSE	Root Mean Squared Error: the standard deviation of regression.
si	ith R-Student or Externally Scaled Studentized Residual.
ri	ith Standardized or Internally Scaled Studentized Residual.
D	Cook's D Influence Diagnostic.
SEBETA	(px1) vector of model coefficient standard errors. where
PRESS	Predicted Error Sum of Squares. PRESS = e'(i)e(i)

For more on PRESS and other displayed statistics, see Linear Model Statistics Displays.

Linear Regression

Prediction Error Variance