Combine columns of multiple arrays using for loop

I am trying to use a for loop to combine columns of multiple arrays in MATLAB.

I want to take column 1 (or i) of the first variable (lat) and combine this with column 1 (or i) of the other three variables (lon, rng, and z), similar to the output expected from the code below.

if true
  % Example of output wanted:
test = [lat(:,1) lon(:,1) rng(:,1) z(:,1)];
end

However, I want to repeat this for all 360 columns in each of the four arrays (see attached matlab file).

How would I do this with a for loop?

[Paolo]

You mean for all 967 columns?

[Charlotte Findlay]

Hi Paolo, it should be 360 columns, but I made a mistake and forgot to adjust the data using the following code:

if true
lat = lat';
lon = lon';
rng = rng';
z = z';
end

[James Tursa]

What would the dimensions of the result 'test' be? Are you wanting to stack these variables vertically or horizontally?

[Charlotte Findlay]

Preferably the output would be vertical with the ability to decipher where one column has ended and another starts (i.e. have another column which labels the 967 rows as 1 and then the next 967 rows as 2, etc.)

Answer 1

[Guillaume]

Preferably the output would be vertical with the ability to decipher where one column has ended and another starts

Is this for storage or display?.

If for storage, then simply concatenate the variables.

allvars = cat(3, lat, lon, rng, z);  %concatenate the variables in the 3rd dimension
%allvars(:, n, :) correspond to column n of the variables

If you wish you can then permute dimensions so that your original variables are then columns,

allvars = permute(cat(3, lat, lon, rng, z), [1 3 2]);
%allvars(:, :, n) corresponds to column n of the variables

edit:

i.e. have another column which labels the first 967 rows as 1 and then the next 967 rows as 2, etc

That can be done as well:

allvars = [reshape(cat(3, lat, lon, rng, z), [], 4), repelem(1:size(lat, 2), size(lat, 1))']

The last column is the index of the original column

[Charlotte Findlay]

Hi Guillaume,

I have edited your above code as follows and it seems to be giving me the output I require in columns:

if true
allvars = permute(cat(3, lat, lon, rng, z), [1 3 2]);
allvars(:, :, 1:360)
end

Would it be possible to add another column to this which denotes 'transect' one i.e. the first 967 rows and 4 columns, and then transect two, etc?

Edit

That additional comment seemed to do just the trick! Thank you so much for your help Guillaume!

Answer 2

[Paolo]

Hi Charlotte,

The following code creates a multidimensional matrix with 967 rows (first dimension), 4 columns (second dimension) and 360 elements (third dimension).

 load('ArrayValuesForLoop_20180521_CFindlay.mat');

 %Dimensions.
 [rows,columns] = size(lat);

 %Number of sheets to combine.
 n = 4;

 %Initialisation.
 output = zeros(rows,n,columns);

 for i = 1:columns
     output(:,:,i) = [lat(:,i) lon(:,i) rng(:,i) z(:,i)];
 end

You can access the output with:

output(:,:,n)

where n is the element you are interested in. As in your example, test will be equal to output(:,:,1).

 test = [lat(:,1) lon(:,1) rng(:,1) z(:,1)];

is equivalent to output(:,:,1).

 test = [lat(:,2) lon(:,2) rng(:,2) z(:,2)];

is equivalent to output(:,:,2), and so on.

You must use isequaln to treat the NaN value as identical.

[Guillaume]

I'm not sure why this answer was deleted to then be reposted more or less the same. I'll repost my comment:

If that is indeed the output wished, then using a loop for that is a complete waste of time:

output = permute(cat(3, lat, lon, rng, z), [1 3 2]);

is all that is required. The above concatenate the 4 arrays in the 3rd dimension, then swap 2nd and 3rd dimension

[Paolo]

Apologies, when I updated the rows and columns with the new values and tested for equality using isequal I was getting a 0 as output, so I thought something was wrong, whereas its just a matter of using isequaln.

Absolutely, its only one of the possible solutions, however Charlotte asked for a solution which uses a for loop in her first post.

[Charlotte Findlay]

Thank you both for your responses. I've tested both and they give me the 3D array.

However, I'll repeat my previous reply above as it would be preferential for the output to be vertical with the ability to decipher where one column has ended and another starts (i.e. have another column which labels the first 967 rows as 1 and then the next 967 rows as 2, etc.)

[Paolo]

Simply increase the second dimension to 5 and store the index in the last column.

output = zeros(rows,5,columns);

for i = 1:columns
    output(:,1:4,i) = [lat(:,i) lon(:,i) rng(:,i) z(:,i)];
    output(:,5,i) = i;
end

[Charlotte Findlay]

Hi Paolo,

Really appreciate your help with this query, and your solution has been made note of in my code! Definitely always multiple solutions to one problem! :)

[Paolo]

You are welcome, there are many ways this can be done. Guillaume's solution is indeed more elegant, I used a for loop in my solution since that was what you asked for.