To get sum values of other columns based on first column values - matlab

Question

An excel file contains 5 columns; first column contains year (1987 to 2080), second column contains month, third column contains days, fourth and fifth column contain values. I would like to get the sum values of columns four and five according to year in column one. For example, I would like to get the sum values of column four and five for year 1987, then 1988, then 1989...so on.!
Example of data file is attached
I have tried the following code considering that each year contains 365 days.
n=1;
for i=1:365:size(data,1)
Total(n,:) = sum(data(i:i+365-1,:));
n=n+1;
end
But the problem is that not all the years contain 365 days. Some of them (e.g. 1988, 1992) contain 366 days in a year as they are leap year. In those cases, the sum results become incorrect.
Looking for your help to get the sum values of columns 4 and 5 according to the year in column 1.
It would be greatly appreciated.

Answer 1

UPDATE: much faster solution at the end!
It can be done as follows with one line for each column:
% some example data
years = ceil(1987:0.3:2080)';
months = randi(12,numel(years),1);
days = randi(30,numel(years),1);
values = randi(42,numel(years),2);
% data similar to yours;
data = [ years months days values ];
That would be the easy readable long way:
% years
y = data(:,1)
% unique years
uy = unique(y);
% for column 4
C4 = arrayfun(#(x) sum( data(y == x, 4) ), uy )
% for column 5
C5 = arrayfun(#(x) sum( data(y == x, 5) ), uy )
or just short in one line per column:
C4 = arrayfun(#(x) sum( data( (data(:,1) == x), 4) ), unique(data(:,1)) )
returning a 94x1 double array with all sums for all 94 unique years of the example data.
If you want to arrange it somehow you could do it as follows:
summary = [uy, C4, C5]
returning something like:
summary = %//sum of sum of
column 4 column 5
1987 3 3
1988 40 40
1989 56 56
1990 96 96
1991 54 54
1992 15 15
1993 73 73
1994 42 42
1995 66 66
1996 56 56
...
You could also do all columns at once. Already for just 2 column it should be 50% faster.
cols = 4:5;
C = cell2mat( arrayfun(#(x) sum( data(y == x, cols),1 ), uy,'uni',0 ) )
The problem with that solution is, that you have a matrix of about 30000x5 size, and for every unique years it will apply the indexing on the whole matrix to "search" for the current year which is summed up. But actually there is an in-built function doing exactly that:
A simpler and much faster solution you can achieve using accumarray:
[~,~, i_uy] = unique(data(:,1));
C4 = accumarray(i_uy,data(:,4));
C5 = accumarray(i_uy,data(:,5));