How to synchronize matrices (timeseries)?

May Somebody help me solve this problem?

I retrieved stock data from Yahoo and stored it in matrices, every matrix consists of 6 Columns (date,open,high,low,close,volume) and N time steps in the rows (daily).

conn = yahoo('http://download.finance.yahoo.com');

DBK=fetch(conn,'DBK.DE',{'Open', 'High', 'Low','Close','Volume'},'Jan 01 2005','Dec 31 2010', 'd')

DTE=fetch(conn,'DTE.DE',{'Open', 'High', 'Low','Close','Volume'},'Jan 01 2005','Dec 31 2010', 'd')

You will notice that both matrices do not have the same amount of rows DBK<1547x6>; DTE<1528x6>

Hence I cannot compare them, since there are some missing values in DTE. But since I got the date ID in column 1 of each matrix I would like to have the intersection between both matrices based on this ID. Can you help me? And is it possible to compute the intersection for more than 2 matrices?

Help is greatly appreciated!

Answer 1

[Léon]

That is exactly what I need, but I have to find the intersection between more than 2 matrices. :-( I tried to use the intersect() command iterative (intersection between A/B, B/C, C/D,…) but this doesn't really work.

Do you have another hint for me, please?

[Fangjun Jiang]

You can do one at a time, like

out=intersect(A,B);

out=intersect(out,C);

out=intersect(out,D);

or see http://www.mathworks.com/matlabcentral/fileexchange/30725-multiple-or-partial-intersect-function

Answer 2

[Fangjun Jiang]

Did you try intersect()?

A=[(1:10)', rand(10,1)];
B=[(5:15)', rand(11,1)];

[C,IndA,IndB]=intersect(A(:,1),B(:,1));

A_Selected=A(IndA,:)
B_Selected=B(IndB,:)

Answer 3

[Léon]

Thank you that was exactly the point!

Answer 4

[Léon]

I have a follow-up question and I hope you can help me again:

I tried to solve this problem using loops and stored the data in cell arrays. Applying your code to that did not quite work, since I managed to get the intersection of all time-series, but now I want to reduce all matrices based upon this intersection:

 % N = number of time-series
  for n = 1:N,

      % Get the intersection of all time-series (date is stored in column 1)
      AnB = intersect(data{n,1}(:,1),data{n,1}(:,1));

end;

for n = 1:N,

      % Now I want to "delete" all rows that are not part of the intersection
      % But of course I want to have all columns with the data in it and not only the date column

      [c, a, b] = intersect(AnB(:,1),data{n,1}(:,1));
      data{n,1} = data{n,1}(a,:); % --> this gives the error

end;

Thank you very much for helping me!

[Fangjun Jiang]

Please post this as a separate question.

And you need to take a second look at the question. None of the variable N, data are given so others won't be able to run the code.

What is AnB = intersect(data{n,1}(:,1),data{n,1}(:,1))?

Do you mean AnB = intersect(data{n,1}(:,1),data{n,2}(:,1))?

[Léon]

OK, sorry for that. I opened a new question: http://www.mathworks.de/matlabcentral/answers/16786-intersection-of-multiple-time-series and reworked the code.