cell to matrix

I have a data set which is in a cell structure (<1x20>) I want to place all of the data into one matrix. As each of the data sets in this cell structure is of different size, such as the first one is: 61169x15 and the second is 59529x15 and so on... i'm trying to find a way of combining the cells into a matrix. I can do this manually by:

new_data=[data1;data2;data3;data4]

But i would like to run the same script for other data sets therefore am looking for a way of doing this in a loop. Is that possible? So, I was thinking of a loop which ran though the cells i.e. for i=1:size(data,2) %=1:20

and then applying the same process as above in the loop, but im not sure on how I would go about doing this.

thanks

[Jan]

Please by careful with the term "structure". It is often confused with "STRUCT" array.

Answer 1

[Jan]

If all cell elements have teh same number of columns:

new_data = cat(1, data{:});

[ricco]

This doesn't seem to work for my data set. I may not be explaining this correctly. So, to begin with I have:

data <1x20 cell>

and then in this I have:

61169x15 ; 59529x15; etc... where all have the same number of columns but different number of rows.

If I select the data, it is described as data{1,1}, data{1,2}...towards data{1,20}.

By typing cell2mat(data') is gives an error that argument dimensions are not consistent, and also gives the same error by typing:

new_data=cat(1,data{:});

any further suggestions?

thanks

[Jan]

@ricco: This is not possible. If all the arrays in the cell have the same number of columns, they can be concatenated using CAT(1, C{:}) or CELL2MAT(C'). If this does not work for your problem, the number of columns must be different. Please check this again:

cellfun('size', data, 2)

[ricco]

It works now, thank you very much for your help. The problem was that the data wasn't cleaned by who ever recovered it so one of the arrays had a different number of columns (14 not 15) due to an instrument malfunction. Which I probably wouldn't have spotted without your comment above.

thanks again

Answer 2

[Fangjun Jiang]

If all cells have the same number of columns, would this help?

a{1}=rand(3,2);
a{2}=rand(4,2);
a{3}=rand(5,2);

cell2mat(a')

[ricco]

In principle it should work because its basically the same as you have shown but for some reason it doesn't. I receive an error:

??? Error using ==> cat

CAT dimensions are not consistent

does it make a difference if mine are 'double' cells?

[Jan]

@ricco: No, the type of the cell elements does not matter - btw.: Fangjun's data are DOUBLEs also. But the dimensions must be matching. While "cell2mat(a')" works, "cell2mat(a)" must fail.

Answer 3

[Ora Zyto]

From your description, it seems like all of your data sets have the same number of columns, and varying number of rows. In your for loop, append your current dataset to your existing matrix:

for i=1:length(data),
new_data=[new_data;data{i}]
end

Ora

[ricco]

Yes the data sets have the same number of columns but varying number of rows, the loop that you noted above doesnt work because new_data has not been defined. Is there anything im missing?

thanks

[Jan]

@Ora Zyto: The loop let the array grow in each iteration. This is very inefficient.

[Ora Zyto]

@Jan Simon: You are totally right. I didn't think about the CAT function :-)