MATLAB: How to select max columns values based on certain column values

differential equationsMATLABmeanstatisticstableuitable

Dear Experiences …
i have table look like the following..
Obs-name      var1    var2   var3 ...... varn      seg_id 
ob1           0.12    0.14   0.17         1.2         1
ob2           1.2     0.2    0.14         0.0         1
ob3           1.5     0.3    1.5          7.2         2
ob4           2.4     4.5    2.2          0.0         3
.......................................etc.
i'm doing the following procedures:
1- i eliminate the first var content .. (obs name), using the following
T = readtable('data.xls');
Tnew=T(:,2:end)
2- calculate averages of all vars values based on segment_ID 
Tavg=varfun(@mean,Tnew,'GroupingVariable','Seg_ID')
– now the tavg table includes all the average values for all vars based on Seg_ID.
Now i need to select the top var values (K) for every Seg_ID (K, where k=5).. then write table includes the Seg_ID , GroupCount , top K vars names look like the following table :
   seg_id(1) Group count (3) var1 name   var4 name    var100 name       
   seg_id(2) group count (5) var2 name   var3 name    var15 name       
   seg_id(3) Group count (10)var1 name   var12 name   varn name         
    ...etc..
attached file include a portion of my data table.
i will thanks any one can give a solution for this issue .. thanks

Best Answer

Not really clear on your question. If you're wanting to calculate the mean of the variables, grouped by seg_ID, then:
varfun(@mean, T, 'InputVariables', 2:width(T)-1, 'GroupingVariables', 'seg_ID')
I'm not sure what you want to do after that
edit: maybe you're looking for this function
function varargout = filterrows(rowvalues, varnames)
%function to be used with rowfun, returns the 5 highest column means of the rows together with their names, interlaced.
%requires 'SeparateInputs', false in the rowfun options.
    rowmeans = mean(rowvalues, 1);
    [~, order] = sort(rowmeans, 'descend');
    varargout = [num2cell(rowmeans(order(1:5))); num2cell(varnames(order(1:5)))];  %will be reshaped into a row vector by rowfun
end
which you can use with rowfun:
rowfun(@(rowvalues) filterrows(rowvalues, T.Properties.VariableNames(2:end-1)), ...
       T, ...
       'InputVariables', 2:width(T)-1, ...
       'GroupingVariables', 'seg_ID', ...
       'SeparateInputs', false, ...
       'NumOutputs', 10)
Not sure it's a good idea, though.
Related Question