How to ensure that MATLAB reads the first column as a column in a table?

coefficient_tab = cell2table(cell(0,5), 'VariableNames', {'Name', 'Estimate', 'SE', 'tStat', 'pValue'});
bundle = ['AB'; 'CD'; 'EF'; 'GH'];

for i = 1:length(bundle) 
        name = bundle(i, :);
        tab = table([value1(i); value2(i)],... % elements of first column
                    [NaN; NaN],... % elements of second column
                    [NaN; NaN],...
                    [NaN; NaN],...
                    'VariableNames',{'Estimate', 'SE', 'tStat', 'pValue'},...
                    'RowNames',{strcat('var1_', name), strcat('var2_', name)});
        coefficient_table = [coefficient_table; tab];
end 

Here's a snapshot of the output from "tab" when i = AB

Estimate SE tStat pValue

________ ___ _____ ______

var1_AB 1.0693 NaN NaN NaN

var2_AB 2.1268 NaN NaN NaN

The output of tab is a table with 4 columns. The columns have values on 'estimate', 'SE', 'tstat', 'pValue'. However, MATLAB is not reading the column with the variable names as a column. So, instead of having 5, there are 4 columns. How can I adjust "tab" to ensure that the output has 5 columns and the first column is named as "Name" as mentioned in "coefficient_tab"?

So, I would like the table to look like:

Name Estimate SE tStat pValue

_____ ________ ___ _____ ______

var1_AB 1.0693 NaN NaN NaN

var2_AB 2.1268 NaN NaN NaN

I tried adding "Name" in "tab", but that didn't work

tab = table([value1(i); value2(i)],... % elements of first column
                    [NaN; NaN],... % elements of second column
                    [NaN; NaN],...
                    [NaN; NaN],...
                    'VariableNames',{'Name','Estimate', 'SE', 'tStat', 'pValue'},...
                    'RowNames',{strcat('var1_', name), strcat('var2_', name)})

Answer 1

[Cris LaPierre]

You are using 'RowNames', which are not considered a table column. You therefore either need to use rownames everywhere (original table is 0x4) or you need to create a 5th variable containing the names and not use rownames.

% Using RowNames
coefficient_table = cell2table(cell(0,4), 'VariableNames', {'Estimate', 'SE', 'tStat', 'pValue'});
bundle = ['AB'; 'CD'; 'EF'; 'GH'];
value1 = rand(1,4);
value2 = rand(1,4);

for i = 1:length(bundle) 
        name = bundle(i, :);
        coefficient_table = [coefficient_table; [{value1(i); value2(i)},... % elements of first column
            {NaN; NaN},... % elements of second column
            {NaN; NaN},...
            {NaN; NaN}]];
end 
coefficient_table.Properties.RowNames = ["var1_";"var2_"] + string(bundle)'

coefficient_table = 8×4 table
               Estimate    SE     tStat    pValue
               ________    ___    _____    ______

    var1_AB    0.058393    NaN     NaN      NaN  
    var2_AB     0.71082    NaN     NaN      NaN  
    var1_CD     0.39971    NaN     NaN      NaN  
    var2_CD     0.65425    NaN     NaN      NaN  
    var1_EF      0.9427    NaN     NaN      NaN  
    var2_EF    0.066835    NaN     NaN      NaN  
    var1_GH     0.14773    NaN     NaN      NaN  
    var2_GH     0.73154    NaN     NaN      NaN  

Now the same code but using a 5th variable instead

coefficient_table = cell2table(cell(0,5), 'VariableNames', {'Name','Estimate', 'SE', 'tStat', 'pValue'});
bundle = ['AB'; 'CD'; 'EF'; 'GH'];
value1 = rand(1,4);
value2 = rand(1,4);

for i = 1:length(bundle) 
        name = bundle(i, :);
        coefficient_table = [coefficient_table; ...
            [{"var1_"+ name;"var2_"+ name},...
            {value1(i); value2(i)},... % elements of first column
            {NaN; NaN},... % elements of second column
            {NaN; NaN},...
            {NaN; NaN}]];
end 
coefficient_table

coefficient_table = 8×5 table
      Name       Estimate    SE     tStat    pValue
    _________    ________    ___    _____    ______

    "var1_AB"    0.48427     NaN     NaN      NaN  
    "var2_AB"    0.94618     NaN     NaN      NaN  
    "var1_CD"     0.3964     NaN     NaN      NaN  
    "var2_CD"     0.8816     NaN     NaN      NaN  
    "var1_EF"    0.41481     NaN     NaN      NaN  
    "var2_EF"    0.46532     NaN     NaN      NaN  
    "var1_GH"    0.43444     NaN     NaN      NaN  
    "var2_GH"    0.23924     NaN     NaN      NaN  

[alphabetagamma]

Thank you for showing the codes as demo. This was helpful

Answer 2

[Walter Roberson]

MATLAB does not put a title on RowNames. If you want a title for that column, you will need to treat it as any other variable.

By the way: if you take an existing table, and you assign

tab(end+1, :) = EXPRESSION

and EXPRESSION is a cell array with as many entries as there are columns, then MATLAB will append the content of the cell entries as a new row in the table. So you could

tab(end+1,:) = {strcat('var1_', name), value1(i), nan, nan, nan};
tab(end+1,:) = {strcat('var2_', name), value2(i), nan, nan, nan};

Answer 3

[dpb]

rownames are table metadata; they are NOT a variable which is why the column variable heading doesn't show up for them -- the table has only four columns of data, not five. The RowNames property of the table returns the values or you can use the following --

tab.(tab.Properties.DimensionNames{1})

See table documentation about RowNames for details.

If you want the names to be a variable, then you have to create the variable that contains them.

You can construct the above table in much more succinct fashion, however --

vnames={'Estimate', 'SE', 'tStat', 'pValue'};
rnames=["Var"+repmat(string([1;2]),numel(bundle)/2,1)+"_"+string(char(kron(bundle,ones(2,1))))];
data=[abs(randn(8,1)) nan(8,3)];
tT=array2table(data,'VariableNames',vnames,'RowNames',rnames);

produces

>> tT
tT =
  8×4 table
               Estimate    SE     tStat    pValue
               ________    ___    _____    ______
    Var1_AB    0.21916     NaN     NaN      NaN  
    Var2_AB     1.0458     NaN     NaN      NaN  
    Var1_CD    0.95098     NaN     NaN      NaN  
    Var2_CD     0.7948     NaN     NaN      NaN  
    Var1_EF    0.07143     NaN     NaN      NaN  
    Var2_EF    0.77369     NaN     NaN      NaN  
    Var1_GH    0.77416     NaN     NaN      NaN  
    Var2_GH    0.26565     NaN     NaN      NaN  
>> 

[alphabetagamma]

I see. Thanks a lot. Although a little complicated, thanks for showing me how to write the code in more succinct format

[dpb]

Seems far less "complicated" than all the machinations through for loop -- and the first creation that is then dynamically reallocated in the for loop is very inefficient. For small tables it won't be particularly noticeable, but presuming the real case is significantly larger the hit may become sizable.

I got just a little cute in the creation of the row names to duplicate the char() string, granted -- "just playing" to give a little "push" to the learning curve; use string arrays with compose() and perhaps an arrayfun() to build the rownames vector in a more straightforward, "dead-ahead" programming style, but the key illustration is to build the data array first and array2table it, don't just add a single record at a time dynamically reallocating the whole table every time through the loop.

There are times when it's about the only way to proceed without a lot of extra effort when one doesn't have the next record yet, but this isn't a case of that; by the example, you already have all the data excepting the unknowns; it would be the case you don't have those until some other calculation that might call for it.

I've submitted enhancement request for an append function to specifically allow one to insert an empty record into an existing table where it is smart enough to know the types and build it -- so far, TMW hasn't seen fit to implement it.

[alphabetagamma]

Oh I see. That explains why my code runs so slowly when the increase the number of elements in "bundle" to 30, instead of just 4. Thanks for the explanation. I'll try to implement it.

[dpb]

Bound to be at least a part of it, anyways, yes...although with still only 60 lines (unless the multipler of number of records/type is also gone up to a higher number than2) I'd not expect it to be all that big of an effect yet.