Cumulative sum by group

Hi, Could someone explain how to compute the running sum by group (without using for loop)? A example is: groupid = [1; 1; 1; 2; 2; 3; 3;]; value=[1;2;3;4;5;6;7]. the desired result is: runsum = [1;3;6; 4;9; 6;13]. Thanks a lot. zppine

Answer 1

[Sean de Wolski]

Using a for-loop here may be your best option. However, my unconditional love for accumarray forces me to present you with this:

groupid = [1; 1; 1; 2; 2; 3; 3;]; 
value=[1;2;3;4;5;6;7]; 
gcs = cell2mat(accumarray(groupid,value,[],@(x){cumsum(x)}))

Answer 2

[zppine]

Hi, Sean. Very good idea. Thanks a lot.

Answer 3

[Daniel]

function S = labeledCumsum(X,L)
% e.g. X=[3   5 8  9 1  2    5    8 5  4  9   2]
%      L=[0   1 1  2 2  0    0    1 1  1  0   5]
% result=[NaN 5 13 9 10 NaN  NaN  8 13 17 NaN 2]

L = L(:);
X = X(:);

if numel(L) ~= numel(X)
    error('The two inputs should be vectors of the same length.')
end

% Do the full cumulative sum
X(isnan(X)) = 0;
S = cumsum(X);

mask = logical(L);

% Lookup the cumulative value just before the start of each segment
isStart = mask & [true ; L(1:end-1) ~= L(2:end)];
startInds = find(isStart);
if startInds(1) == 1
    S_starts = [0 ; S(startInds(2:end)-1)];
else
    S_starts = S(startInds-1);
end

% Subtract off the excess values (i.e. the ones obtained above)
L_safe = cumsum(isStart); % we do this to main case the labels in L were not sorted integers
S(mask) = S(mask) - S_starts(L_safe(mask));

% Put NaNs in the false blocks
S(L==0) = NaN; 
end

Answer 4

[Quoc T. Phan]

Hello

When it comes to work with group, these two functions are great tools. They go together

1. findgroups: --> indexing groups
2. splitapply: --> split your data by index, then apply @function to it, then combine back to original

Here is a sample code

clear
load patients
% Create a table X
X = table(LastName,Gender,Height,Weight,Age)

% Sort rows by Gender and Last Name (Actually u dont need Last Name, but I
% just put here in case)
X =  sortrows(X,{'Gender' 'LastName'},{'ascend' 'ascend'})

% Function 01: Using findgroups --> to return logical index
G = findgroups(X.Gender) 

% (*) Note: by default they index by alphabet order. Although 1st obs is male,
% however, F(emale) is before M(ale). Then 1 is Female and 2 is Male.


% Function 02: Using splitapply
% Basically what it does it slit table X by logical index G
% It then apply function cumsum for variable X.Height
cumHeight = splitapply(@(x1){cumsum(x1)},X.Height,G)

% Change cell to matrix and add to table X
X.cumHeight = cell2mat(cumHeight)