Sorting columns in classes

Hello I have table

 T = 

    Distance       ddfi       ddla 
    _________    ________    ______

      0.25445    0.032528    330.13
      0.11412    0.023379    330.85
      0.10778    0.022379     330.9
      0.12522    0.022749    331.04
      0.11136    0.018143    331.56
       0.1489     0.02429    331.24
       0.2754     0.03361    330.62
      0.13454    0.017196    331.82
       0.1302    0.018947    331.64
      0.33098    0.037232    330.54
      0.15657    0.023177    331.48
       0.2157    0.029463    331.39
      0.15954    0.019613     331.9
      0.17042    0.021007    331.92
       0.1996     0.02608    331.59
      0.18505    0.019299    332.16
      0.20692    0.024783    331.71
      0.21569    0.026374    331.68
      0.19809    0.021438    332.14
      0.26943    0.030838    331.39

So I need to sort this table by first column, but in a specific interval so to get classes. For example: I sort table by first row using interval 0.05 to get classes and the sum all data in same class from second column and third column. This is done in excel using pivot table and grouping:

Row Labels  Sum of ddfi  Sum of ddla
0.10778-0.15778  0.170  2650.530
0.15778-0.20778  0.067  995.410
0.20778-0.25778  0.062  661.520
0.25778-0.30778  0.034  330.620
0.30778-0.35778  0.037  330.540
Grand Total  0.369793  4968.62

Thanks for help

Answer 1

[Guillaume]

A simpler option than splitapply is to use varfun:

edges = min(T.Distance):0.05:max(T.Distance)+0.05;
group = discretize(T.Distance, edges);
groupnames = compose('%f-%f', edges(1:end-1)', edges(2:end)');
T.rowlabels = groupnames(group);
newtable = varfun(@sum, T, 'GroupingVariables', 'rowlabels', 'InputVariables', {'ddfi', 'ddla'})

[giometar]

Got this error:

Undefined function 'compose' for input arguments of type 'char'.

Error in Modeliranje_kovarijance (line 47)
groupnames = compose('%f-%f', edges(1:end-1)', edges(2:end )');

Error in run (line 96)
evalin('caller', [script ';']);

[Guillaume]

Undefined function 'compose'

That's because you're on a version of matlab earlier than R2016b. It's always a good idea to say that you're on an old version in your question (or fill the Release field next to the question).

On old versions, you can replace compose by the undocumented sprintfc:

groupnames = sprinfc('%f-%f', [edges(1:end-1); edges(2:end)]');

or use a loop:

groupnames = arrayfun(@(s, e) sprintf('%f-%f', s, e), edges(1:end-1), edges(2:end), 'UniformOutput', false);

or not bother with the groupnames (which is only for pretty display of the groups) and do:

T.rowlabels = group;

instead.

[giometar]

Sorry. My mistake. It works fine...

Answer 2

[Matt J]

You don't need sorting to do that. Just use discretize()

 G=discretize(T{:,1},edges);

then use splitapply on each of the variables

for i=[3,2]
 result(:,i-1)=splitapply(@sum,T{:,i},G);
end

 newTable=array2table(result,'VariableNames',T(:,2:3).VariableNames);

[Guillaume]

giometar's comment mistakenly posted as an answer moved here:

It doesn't work. Probably I am doing something wrong. My table is stored using array2table function. Result is table T (as in first post). When I use function discretize I got an error: "Undefined function or variable 'edges'." Ok I put edges like [0 0.05] and then got result for G. After that for loop finished but next step is error again: "You cannot access the 'VariableNames' property directly. Access it using dot subscripting via.Properties.VariableNames." What am I doing wrong?

[Guillaume]

Matt made a mistake.

T(:,2:3).VariableNames

should be

T.Properties.VariableNames(2:3)

[giometar]

With that code

Table=[D, ddfi, ddla];
T = array2table(Table,...
    'VariableNames',{'Distance','ddfi','ddla'});
G=discretize(T{:,1},[0 0.05]);
for i=[3,2]
 result(:,i-1)=splitapply(@sum,T{:,i},G);
end

newTable=array2table(result,...
    'VariableNames',T.Properties.VariableNames(2:3));

I get empty 0x2 newTable