CookD

As of Version 7.0, CookD has been renamed to "CookDistances" and has become a property of LinearModelFit.

CookD

is a possible value for the RegressionReport option for Regress and DesignedRegress which represents a list of Cook's distances.

Details and Options

To use CookD, you first need to load the Regression Common Functions Package. You can do this by using Needs["LinearRegression`"], which will automatically load the Regression Common Functions Package, or you can load the package directly by using Needs["RegressionCommon`"].
CookD is a diagnostic for identifying highly influential points.
CookD gives a list of Cook's distances for the data points.
CookD is equivalent to HatDiagonal/(1-HatDiagonal) StandardizedResiduals²/p where p is the number of parameters in the model.

Wolfram Language code: Needs["LinearRegression`"]

Sample data:

Wolfram Language code:

data = {{0.05, 90}, {0.09, 95}, {0.14, 110}, 
{0.17, 125}, {0.2, 140}, {0.21, 150}, 
{0.23, 175}, {0.25, 190}, {0.3, 210}, 
{0.35, 255}};

CookD for a linear regression:

Wolfram Language code: Regress[data, {1, x ^ 2}, x, RegressionReport -> {CookD}]

Top