How can I average a 4-D array every nth element without using too much loops

Hi,

So I am currently working on a (476,238,1,2699) array using MatLab 2017a corresponding to (longitude, latitude, 1, time). I basically want to make a monthly average of the data (so every 5th point). The problem is I tried using a loop but it takes so much time. Here is the code I am using :

data_zeros = [];
for i = 1:2699
  for j = 1:238
    for k = 1:476
      if (mod(i, 5) == 0)
        data_zeros(k,j,1,i) = mean(data(k,j,1,i-4:i));
                  i
                  j
                  k
      end
    end
  end
end

I need your insight, please help :) !

[dpb]

If you have data by time for lat,long, why store in a 4D array--why not use a time table or such to make the time-based operations simpler?

What form is the time data in?

[Isma Di Carlo]

The data is in seconds starting from 15th January 1947. I have a time table. Like I know a point corresponds to that point in time. I just need to make a monthly average and I estimated that 5 points of data = 1 month.

(The 4D array is because the 3rd object is the depth. I just only took the 10 first meters of my dataset)

[Walter Roberson]

"I have a time table."

Use retime()

Answer 1

[OCDER]

data = rand(476, 238, 1, 2699);
N = floor(2699/5);
data_zeros = zeros(476, 238, 1, N); %Preallocate this for speed
k = 1;
for i = 5:5:2699
    data_zeros(:, :, 1, k) = mean(data(:, :, 1, i-4:i), 4); %Mean along 4th dimension
    k = k+1;
end

[Isma Di Carlo]

This works perfectly ! Thanks !

Answer 2

[Walter Roberson]

mean( reshape(data, size(data,1), size(data,2), 5, size(data,4)/5 ), 3 )

I take advantage here of the fact that the third dimension is length 1 to pull in groups of 5 into the third dimension and mean() along that dimension.

However, you have the difficulty that your data is not an exact multiple of 5 long in that dimension. How do you want to take the mean of the entries 2696:2699 ?

[Isma Di Carlo]

Works amazingly like the other answer ! I interpolated for the last data point !