isnan to conver nan to zero in a table

Hi All,

Apologies for what might be a silly question but I am stuck at the moment.

I have used readtable to input a text formatted CSV file.

T=readtable(zz,'Delimiter', '\t');

This gives a 39x501 table.

The table has NaN values which I need to replace with zeros. Based on anoher answer, I am trying:

T(isnan(T)) = 0;

but get the error

Incorrect number or types of inputs or outputs for function isnan.

Help :-)

Darren

Answer 1

[Steven Lord]

Let's make a sample table with some NaN values and a non-numeric variable.

A = magic(4);
A(randperm(numel(A), 3)) = NaN;
T = array2table(A);
T.A5 = ["apple"; "banana"; "cherry"; "durian"]

T = 4x5 table
    A1     A2     A3     A4       A5   
    ___    ___    ___    __    ________

    NaN      2      3    13    "apple" 
      5     11     10     8    "banana"
      9    NaN    NaN    12    "cherry"
      4     14     15     1    "durian"

You can use the fillmissing function to replace the NaN values in the numeric variables with the constant 0.

T2 = fillmissing(T, 'constant', 0, 'DataVariables', @isnumeric)

T2 = 4x5 table
    A1    A2    A3    A4       A5   
    __    __    __    __    ________

    0      2     3    13    "apple" 
    5     11    10     8    "banana"
    9      0     0    12    "cherry"
    4     14    15     1    "durian"

If all the variables are numeric you can omit the 'DataVariables' name-value argument.

[Darren]

Thank you very much, that works absolutely fine :-)

All the best,

Darren

Answer 2

[Fangjun Jiang]

T is an object of the class "table". You can't apply isnan() on it directly. Try the 'TreatAsMissing' option of the readtable() function to avoid NaN first hand.

[Walter Roberson]

TreatAsMissing gives additional data patterns that are to have NaN (or appropriate missing indicator) substituted. TreatAsMissing does not give any opportunity to substitute a constant value for missing data.