Is there something similar as splitapply that works on a data table instead just on one value?

I created a subset of my input data that has more then 400 data sets and called it tex. The Data in each dataset is 13269x3 table, see picture to the right.

I would like to build the maximum for each row of P and QP for all groups of K and L

[perKL,KL,LK] = findgroups(tex.K, tex.L); % The findgroups does its job perferct.

maxByKL = splitapply(@max,tex.Data{:,1}, perKL); % here is the point of question,

Splitapply doesn't work since it expects the same number of rows for the data variable.

Is there already any existing function for that?

Answer 1

[Matt J]

Tex=varfun(@(x){table2array(x)},tex,'InputVariables','Data');

out = varfun(@(z)max(cell2mat(z)), Tex,'InputVariables','Data','GroupingVariables',{'K','L'})

[Désirée Kroner]

Hi Matt J

Thanks for your quick answer.

I get the following Error on the first line.

Tex=varfun(@(x){table2array(x)},tex,'InputVariables','Data');

[Matt J]

It would be easier to troubleshoot if you Run it here in the forum, rather than on your own computer.

[Désirée Kroner]

Thanks for pointing out. It is my first time asking a question in the forum.

load('tex.mat','tex')
tex

tex = 12×4 table
    K    L         Data          M
    _    _    _______________    _

    E    X    {13269×3 table}    8
    E    Y    {13269×3 table}    8
    E    Z    {13269×3 table}    8
    B    X    {13269×3 table}    8
    B    Y    {13269×3 table}    8
    B    Z    {13269×3 table}    8
    B    X    {13269×3 table}    7
    B    Y    {13269×3 table}    7
    B    Z    {13269×3 table}    7
    E    X    {13269×3 table}    7
    E    Y    {13269×3 table}    7
    E    Z    {13269×3 table}    7

[perKL,KL,LK] = findgroups(tex.K, tex.L);
Tex = varfun(@(x){table2array(x)},tex,'InputVariables','Data');

Error using tabular/varfun>dfltErrHandler (line 436)
Applying the function '@(x){table2array(x)}' to the variable 'Data' generated the following error:

Input argument must be a table.


Error in tabular/varfun>@(s,varargin)dfltErrHandler(grouped,funName,s,varargin{:}) (line 185)
errHandler = @(s,varargin) dfltErrHandler(grouped,funName,s,varargin{:});

Error in tabular/varfun (line 355)
b_data{jvar} = errHandler(s,a_data{jj});

out = varfun(@(z)max(cell2mat(z)), Tex,'InputVariables','Data','GroupingVariables',{'K','L'})

[Matt J]

load(websave('tex','https://www.mathworks.com/matlabcentral/answers/uploaded_files/894385/tex.mat'));

Tex=rowfun(@(x){table2array(x)},tex,'InputVar','Data','OutputVar','Data','Extract',1);

Tex=[tex(:,1:2),Tex];

out = varfun(@(z)max(cell2mat(z)), Tex,'InputVariables','Data','GroupingVariables',{'K','L'});
 out(:,3)=[];
 out.Properties.VariableNames{3}='maxByKL'

out = 6×3 table
    K    L            maxByKL        
    _    _    _______________________

    B    X    3e+07    15.96    14.18
    B    Y    3e+07    22.75    22.75
    B    Z    3e+07    32.53    31.21
    E    X    3e+07    26.62    25.31
    E    Y    3e+07    16.64    15.39
    E    Z    3e+07    21.97     21.8

[Désirée Kroner]

Thanks so far for your help.

Maybe I was not that clear in my description. I need the max of each Dataset row.

load(websave('tex','https://www.mathworks.com/matlabcentral/answers/uploaded_files/894385/tex.mat'));
tex;
[perKL,KL,LK] = findgroups(tex.K, tex.L);
NrGroupsKL = max(perKL);
rowData = size(tex.Data{1,1},1);

    %Preallocate out table
    out = table(KL,LK);
    out = renamevars(out,["KL","LK"],["K","L"]);
    out.PMax = cell(length(KL),1);
    out.QPMax = cell(length(KL),1);
    out 

out = 6×4 table
    K    L        PMax           QPMax    
    _    _    ____________    ____________

    B    X    {0×0 double}    {0×0 double}
    B    Y    {0×0 double}    {0×0 double}
    B    Z    {0×0 double}    {0×0 double}
    E    X    {0×0 double}    {0×0 double}
    E    Y    {0×0 double}    {0×0 double}
    E    Z    {0×0 double}    {0×0 double}

   
    
  for j = 1:NrGroupsKL
    t   = tex(perKL==j,"Data");
    
    % Perallocate Tex_P and Tex_QP
    col = size(t,1);
    row = size(t.Data{1,1},1);
    Tex_P = zeros(rowData,col);
    Tex_QP= zeros(rowData,col);
    
    % extract Data from current group
    for i = 1:col
        Tex_P(:,i) = t.Data{i,1}.PeakdBAm;
        Tex_QP(:,i) = t.Data{i,1}.QuasiPeakdBAm;
    end
    % fill out table group vice with max for each row
    out.PMax(j) = {max(Tex_P,[],2)};
    out.QPMax(j) = {max(Tex_QP,[],2)};
     
  end
out

out = 6×4 table
    K    L          PMax               QPMax      
    _    _    ________________    ________________

    B    X    {13269×1 double}    {13269×1 double}
    B    Y    {13269×1 double}    {13269×1 double}
    B    Z    {13269×1 double}    {13269×1 double}
    E    X    {13269×1 double}    {13269×1 double}
    E    Y    {13269×1 double}    {13269×1 double}
    E    Z    {13269×1 double}    {13269×1 double}

But I manage it only with a for loop. Which will take alot of time when running the full data set.

[Matt J]

I think this is what you mean,

Tex=rowfun(@(x){table2array(x)},tex,'InputVar','Data','OutputVar','Data','Extract',1);

Tex=[tex(:,1:2),Tex];

out = varfun(@(z){max(cat(3,z{:}),[],3)}, Tex,'InputVariables','Data','GroupingVariables',{'K','L'});
 out(:,3)=[];
 out.Properties.VariableNames{3}='maxByKL'

[Désirée Kroner]

Thank you Matt

This is what I was searching for.