Ridgecv: Difference between revisions

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search
imported>Jeremy
(Importing text file)
 
imported>Jeremy
 
(3 intermediate revisions by 2 users not shown)
Line 1: Line 1:
===Purpose===
===Purpose===
Ridge regression with cross validation.
Ridge regression with cross validation.
===Synopsis===
===Synopsis===
:[b,theta,cumpress] = ridge(x,y,thetamax,divs,split)
 
:[b,theta,cumpress] = ridgecv(x,y,thetamax,divs,split)
 
===Description===
===Description===
The function ridgecv uses cross-validation to create a ridge regression model for a matrix of predictor variables (x-block) x, and a matrix of predicted variables (y-block) y. The maximum value of the ridge parameter to consider is given by thetamax (0 < thetamax). divs specifies the number of values of the ridge parameter between 0 and thetamax to be used for calculating models used in the cross validation and shown in plots created by the routine, and split is the number of times the model is rebuilt on a different subset of samples.
 
Outputs are b the regression column vector at optimum ridge parameter theta as determined by cross-validation.
This function calculates a ridge regression model using a matching set of predictor variables (x-block) <tt>x</tt> and predicted variables (y-block) <tt>y</tt>, and uses cross-validation to determine the optimum value of the ridge parameter <tt>theta</tt>. The maximum value of the ridge parameter to consider is given by <tt>thetamax</tt> (where 0 < thetamax).  
In most instances the optimum ridge parameter will be less than 0.1, often as low as 0.01. A good starting guess when working with the method is to specify thetamax = 0.1 with divs = 20.
 
Note: RIDGECV uses the venetian blinds cross-validation method.
====Inputs====
 
* '''x''' = matrix of independent variables
* '''y''' = matching vector of dependent variables
* '''thetamax''' = maximum value of ridge parameter <tt>theta</tt> to consider
* '''divs''' = the number of values of <tt>theta</tt> to test
* '''split''' =  the number of times to split and test the data for cross-validation
 
====Outputs====
 
* '''b''' = the regression column vector, at the optimal ridge parameter value
* '''theta''' = the optimal ridge parameter value
* '''cumpress''' = Predicted Residual Sum of Squares (PRESS) statistics for the cross-validation
 
'''Note:''' RIDGECV uses the venetian blinds cross-validation method.
 
===See Also===
===See Also===
[[crossval]], [[pcr]], [[pls]], [[analysis]], [[ridge]]
[[crossval]], [[pcr]], [[pls]], [[analysis]], [[ridge]]

Latest revision as of 15:55, 22 February 2013

Purpose

Ridge regression with cross validation.

Synopsis

[b,theta,cumpress] = ridgecv(x,y,thetamax,divs,split)

Description

This function calculates a ridge regression model using a matching set of predictor variables (x-block) x and predicted variables (y-block) y, and uses cross-validation to determine the optimum value of the ridge parameter theta. The maximum value of the ridge parameter to consider is given by thetamax (where 0 < thetamax).

Inputs

  • x = matrix of independent variables
  • y = matching vector of dependent variables
  • thetamax = maximum value of ridge parameter theta to consider
  • divs = the number of values of theta to test
  • split = the number of times to split and test the data for cross-validation

Outputs

  • b = the regression column vector, at the optimal ridge parameter value
  • theta = the optimal ridge parameter value
  • cumpress = Predicted Residual Sum of Squares (PRESS) statistics for the cross-validation

Note: RIDGECV uses the venetian blinds cross-validation method.

See Also

crossval, pcr, pls, analysis, ridge

Retrieved from "https://www.wiki.eigenvector.com/index.php?title=Ridgecv&oldid=2396"