MISVAL

Fixed mapping handling the missing values in a dataset

     B = MISVAL(A,VAL,K)
     B = A*MISVAL([],VAL,K)
     B = A*MISVAL(VAL,K)
     N = A*MISVAL([],K)

Input
A	Dataset, containing NaNs (missing values).
VAL	String with substitution option or value used for substitution.  By default all objects with missing values are removed.
K	Vector with indices of features that should be handled.  Default: all

Output
B	Dataset with NaNs substituted
N	Vector with number of missing values per feature

Description

The following values for VAL are possible
'remove' remove objects (rows) that contain missing values (default)  'f-remove' remove features (columns) that contain missing values  'mean' fill the entries with the mean of the features  'c-mean' fill the entries with the class mean of the features  'median' fill the entries with the median of the features  'c-median' fill the entries with the class median of the features  'majority' fill the entries with the majority of the features  'c-majority' fill the entries with the class majority of the features  fill the entries with a fixed constant  NaN no change: return dataset as it is

Note that replacing missing values by feature means (medians or  majorities) is a very simple solution that might be far from optimal.

See also

datasets, feattypes, ismisval,