How can I re-code numeric variables in a table to strings?

I am using Matlab version 9.1.0.441655 (R2016).

I have created a large (2496000x8) table using array2table. Although it was simpler to create the initial matrix as uniformly numeric, for analysis I would like re-code the numbers in several variables to be strings, while leaving other variable as numeric.

An example data set:

    Var1=[1; 2; 1; 2; 2; 1; 1];
    Var2=[2; 2; 1; 1; 1; 2; 1];
    Var3=[32.1; 34.6; 56.7; 65.3; 78.6; 23.4; 43.1];
    Mat=[Var1,Var2,Var3];

    Table=array2table(Mat,'VariableNames',{'Var1','Var2','Var3'});

Table = 

    Var1    Var2    Var3
    ____    ____    ____

    1       2       32.1
    2       2       34.6
    1       1       56.7
    2       1       65.3
    2       1       78.6
    1       2       23.4
    1       1       43.1

I would like to re-code the 1's and 2's in Var1 to 'Male' and 'Female', respectively and re-code the 1's and 2's in Var2 to 'Old' and 'Young', respectively. But, leave Var3 as numeric.

To result in:

      Var1       Var2        Var3   
    ________    _______    _________

    'Male'      'Young'    [32.1000]
    'Female'    'Young'    [34.6000]
    'Male'      'Old'      [56.7000]
    'Female'    'Old'      [65.3000]
    'Female'    'Old'      [78.6000]
    'Male'      'Old'      [23.4000]
    'Male'      'Young'    [43.1000]

I tried:

Table.Var1(strcmp(Table.Var1,1)) = {'Male'}

But that gives a conversion error:

Conversion to double from cell is not possible.

Answer 1

[Peter Perkins]

JLC, you say "string", but it's very likely that you would be bettter off with categorical variables. Given you're original sample data ...

>> Var1 = [1; 2; 1; 2; 2; 1; 1];
>> Var2 = [2; 2; 1; 1; 1; 2; 1];
>> Var3 = [32.1; 34.6; 56.7; 65.3; 78.6; 23.4; 43.1];
>> t = table(Var1,Var2,Var3)
t =
  7×3 table
    Var1    Var2    Var3
    ____    ____    ____
     1       2      32.1
     2       2      34.6
     1       1      56.7
     2       1      65.3
     2       1      78.6
     1       2      23.4
     1       1      43.1
... here's what you probably want to do:

>> t.Var1 = categorical(t.Var1,1:2,{'Male' 'Female'});
>> t.Var2 = categorical(t.Var2,1:2,{'Old' 'Young'});
>> t
t =
  7×3 table
     Var1     Var2     Var3
    ______    _____    ____
    Male      Young    32.1
    Female    Young    34.6
    Male      Old      56.7
    Female    Old      65.3
    Female    Old      78.6
    Male      Young    23.4
    Male      Old      43.1

A few other things:

• Don't name your table "Table". It will come back to bite you eventually.
• No reason to concatenate [Var1,Var2,Var3] and then call array2table, which just pulls them apart. Call table directly.
• Your "desired" result seems to have Var3 as a cell array, each cell containing a scalar double, You definitely don't want that.
• It's possible to convert the entire table to a cell array and work on it and convert it back, as shown by ImageAnalyst. But you can work on the table directly as I showed above. The "Conversion to double from cell is not possible." errors you saw are because, as IA says, you were assigning text to only some elements of a double variable in the table. You need to assign to t.Var1 without subscripting. My code above shows that as part of converting to categorical.
• If you really do want strings, do it like this:

genders = {'Male'; 'Female'}; t.Var1 = genders(t.Var1);
ages = {'Male'; 'Female'}; t.Var2 = ages(t.Var2);

[JLC]

Thank you very much! Using categorical worked perfectly with my real data.

Answer 2

[Andrei Bobrov]

c = {'Male','Female','Old','Young'};
T_out = cell2table(c(cell2mat(Table{:,:})+[0 1]));

or for the latest versions of the MATLAB:

Table(:,1:2) = array2table(categorical(Table{:,1:2}+[0 1],1:4,{'Male','Female','Old','Young'}));

or with new conditions

Var1=[1; 2; 1; 2; 2; 1; 1];
Var2=[2; 2; 1; 1; 1; 2; 1];
Var3=[32.1; 34.6; 56.7; 65.3; 78.6; 23.4; 43.1];
Table=table(Var1,Var2,Var3);

c = {'Male','Female','Old','Young'};
T_out = [cell2table(c(bsxfun(@plus,Table{:,1:2},0:1))),Table(:,3)];

[JLC]

I tried this with the update more representative example data:

T_out = array2table(categorical(cell2mat(Table{:,1:2})+[0 1],1:4,{'Male','Female','Old','Young'}));

But I get the following error:

Cell contents reference from a non-cell array object.

Error in cell2mat (line 42)
cellclass = class(c{1});

[Andrei Bobrov]

Please see third part "with new conditions" of my answer.

Answer 3

[Image Analyst]

Try this:

Var1={1,2,1,2,2,1,1};
Var2={2,2,1,1,1,2,1};
Mat=[Var1',Var2'];
Table=array2table(Mat,'VariableNames',{'Var1','Var2'})
% Find where column 1 = 1.
rowsWith1 = cell2mat(Table.Var1) == 1
% Replace those rows with 'Male' and other rows with 'Female'.
Table(rowsWith1, 'Var1') = {'Male'}
Table(~rowsWith1, 'Var1') = {'Female'}

% Find where column 1 = 1.
rowsWith1 = cell2mat(Table.Var2) == 1
% Replace those rows with 'Young' and other rows with 'Old'.
Table(rowsWith1, 'Var2') = {'Young'}
Table(~rowsWith1, 'Var2') = {'Old'}

[JLC]

This works well for the example data. Unfortunately my real data is a numeric (double) table, so this won't work. I get an error when I try to re-code from the logical vector:

Conversion to double from cell is not possible.

I have updated the original question with a closer example data set.

[Image Analyst]

JLC:

The problem is that with tables, all elements in a column must be of the same type. So if you started with numbers, you can't then set some elements in some rows to 'Male' which is a string because then the column would contain both numbers and strings, which is not allowed with tables. You CAN to that with cell arrays however. So I converted your table to a cell array, then made the replacements, and then (now that they're all strings) converted back to a table. See this:

Var1=[1; 2; 1; 2; 2; 1; 1];
Var2=[2; 2; 1; 1; 1; 2; 1];
Var3=[32.1; 34.6; 56.7; 65.3; 78.6; 23.4; 43.1];
Mat=[Var1,Var2,Var3];

Table=array2table(Mat,'VariableNames',{'Var1','Var2','Var3'})
% Need to create a cell array, because with a table we can't have some rows in the Var1 column
% be numbers while others are strings.  However a cell array can do that.
ca = table2cell(Table)
% Find where column 1 = 1.
rowsWith1 = Table{:, 'Var1'} == 1
% Replace those rows with 'Male' and other rows with 'Female'.
ca(rowsWith1, 1) = {'Male'}
ca(~rowsWith1, 1) = {'Female'}

% Find where column 1 = 1.
rowsWith1 = Table{:, 'Var2'} == 1
% Replace those rows with 'Young' and other rows with 'Old'.
ca(rowsWith1, 2) = {'Young'}
ca(~rowsWith1, 2) = {'Old'}

% Now convert from cell array back to table.
Table = cell2table(ca, 'VariableNames',{'Var1','Var2','Var3'})