repartition

Repartition data for cross-validation

Syntax

cnew = repartition(c)

cnew = repartition(c,s)

Description

cnew = repartition(c) creates a cvpartition object cnew that defines a random partition of the same type as c, where c is also a cvpartition object. That is, repartition takes the same observations in c and repartitions them into new training and test sets.

cnew = repartition(c,s) uses the RandStream object s as the random number generator for the new partition.

Examples

Repartition Data for Cross-Validation

Repartition observations in a cvpartition object. The type of validation partition remains the same.

Partition 100 observations for 3-fold cross-validation.

c = cvpartition(100,'KFold',3)

c = 
K-fold cross validation partition
   NumObservations: 100
       NumTestSets: 3
         TrainSize: 67  66  67
          TestSize: 33  34  33

Repartition the observations.

cnew = repartition(c)

cnew = 
K-fold cross validation partition
   NumObservations: 100
       NumTestSets: 3
         TrainSize: 67  66  67
          TestSize: 33  34  33

Notice that the set of observations in the first test set (fold) of c is not the same as the set of observations in the first test set of cnew.

isequal(test(c,1),test(cnew,1))

ans = logical
   0

View the validation partition type of c and cnew. Both c and cnew are validation partitions of the same type, 'kfold'.

isequal(c.Type,cnew.Type)

ans = logical
   1

c.Type

ans = 
'kfold'

Input Arguments

c — Validation partition

cvpartition object

Validation partition, specified as a cvpartition object. The validation partition type of c, c.Type, is the same as the validation partition type of the new partition cnew.

s — Random number generator

RandStream object

Random number generator for the new partition, specified as a RandStream object.

Tips

• Repartitioning is useful for Monte Carlo repetitions of cross-validation analyses. crossval calls repartition when you specify the 'MCReps' name-value pair argument.