How do I sum over one dimension of a multidimensional array?

I have four dimensions in a netcdf file: lat, long, sector and time. I want to sum over JUST sector. Is there a simple command to do this?

Answer 1

[Adam Danz]

Are you using ncread() or something similar to load the variable into the matlab workspace? If the variable is a matrix, you can use the 2nd input to sum() to sum across a specified dimension.

Example:

 r = rand(5,4,3,2); %4D data
 squeeze(sum(r,3)) %sums across the 3rd dim.

Squeeze() just removes the extra singleton dimensions.

[Nesha Wright]

I think this would work. Is there a way to tell what order the dimensions are in?

[Walter Roberson]

ncinfo can give you information about the order the variables are in the file. But be careful because nc files are "row major" instead of "column major" -- that is, the memory storage varies most quickly across rows instead of down columns. Because of this it is common that you need to use permute() on the result of ncread() to get the data order you expect. For example,

TheArray = permute(TheArray, [2 1 3 4]);

When you are working with ncf data sets it is useful to pay close attention to the array sizes you get back compared to the length of the marginal index, to determine whether your first two dimensions need to be exchanged or not.

Answer 2

[Juan Fernandez]

Yes, there is a simple single command. Use the sum(A,dim) function. If this is not working, then provide additional information and perhaps I can be of more assistance.

[Nesha Wright]

dimsum=sum(total,3); %sum of the Sector dimension so result is 4-D single.

I need my answer to not be in 4D... How do you remove the other dimensions?

[Walter Roberson]

permute(dimsum, [1 2 4 3])

or

squeeze(dimsum)

[Louis-Philippe Guinard]

Tracing back to Juan's answer, if I wanted to sum, say, just the N first elements of the 3rd dimension, how would one go about that?

I have a very similar problem, as in, a 4D array. I treat it like a 256x256 image at 256 different timestamps, for 30 acquisitions. In this regard, my dimensions would be x, y, t and acq. I want to add the 15 first t's while leaving all other dimensions intact, I don't care what happens to the rest of the t's. How would that be possible?

Thanks in advance!

[Adam Danz]

> if I wanted to sum, say, just the N first elements of the 3rd dimension, how would one go about that?

sum(x(:,:,1:N), 3)

[Louis-Philippe Guinard]

I wasn't too far off, I did sum(x(:,:,1:N)) and returned a 1x256xN array. Aight thanks a lot! :)