MATLAB: Replace NaN values with values based on same group and other group

dataMATLABnantable

  • table (650,000×20)
  • customerID group
  • cost group (NaN values throughout)
The table contains 5-7 data lines for each customerID that have the same cost except for one or two NaN cost values. It is obvious to replace the NaN values with the same other given cost values associated with the ID. But I am having trouble.
I was thinking:
Table.cost(isnan(Table.cost)) = Table.customerID(mode(Train.cost));
But this would apply to the entire rather than correct customerID?

Best Answer

I think this should work.
 % Sample data
 T = cell2table({'ID1' 100 'aaa';  
                  'ID1' nan 'bbb';
                  'ID1' nan 'ccc';
                  'ID2' 200 'ddd';
                  'ID2' nan 'eee';},'VariableNames',{'customerID','cost','otherStuff'})
% Replace the NaNs with the mode of that group          
G = findgroups(T.customerID);
val = splitapply(@(x){mode(x)*ones(size(x))},T.cost,G);
T.cost = cat(1,val{:}) % <----- Fixed now
Related Question