Stratifiedcvi

Purpose

Create a stratified cross-validation index vector for a given method and class set.

Synopsis

cvi = stratifiedcvi(dso,classset,cvinfo)

cvi = stratifiedcvi(dso,1,{'loo'}) using Leave-One out cross-val

cvi = stratifiedcvi(dso,1,{'vet' 5 2}) using venetian blinds cross-val

cvi = stratifiedcvi(dso,1,{'con' 2}) using contiguous blocks cross-val

cvi = stratifiedcvi(dso,1,{'rnd' 2}) using random cross-val

Output (cvi) is a vector containing the group number of each item.

Description

Create a stratified cross-validation index vector for given class set in a Dataset Object and a given cross-val method. This method will apply the cross-val method on each class at a time and then combine the resulting cross-val index vectors.

Inputs

• dso = DataSet Object with class information

• classset = class set to use or vector of numerical class assignments

• cvinfo = cross-validation information, a cell array which should contain:
  • method: a string defining the cross-validation method defined below, can be any of the following:
    • 'vet'  : Venetian blinds. Every n-th item is grouped together. Optionally allows grouping of more than one sample together using the "blocksize" input.
    • 'con'  : Contiguous blocks. Consecutive items are put into n groups.
    • 'loo'  : Leave one out. Each item is in an individual group, input (n) can be omitted.
    • 'rnd'  : Random. items are randomly split into n equal sized groups.
  • n: the number of subsets to split the data into,
  • blocksize: the number of items to include in each block (NOTE: blocksize for 'vet' method only)

Outputs

• outputs = vector of integer values giving the group number of each item.

See Also

crossval, cvifromclass