Question

In MATLAB, how could I combine two matrices of data measured at different frequencies such that the result is indexed at the higher frequency? Since the data measured at the lower frequency will have many unknown values in the result, I would like to replace them with the last known value in the matrix. There is a lot of data so a vectorized solution would be preferred. I’ve added some sample data below.

Given:

    index1  data1  index2  data2
    1       2.1    2       30.5
    2       3.3    6       32.0
    3       3.5    9       35.0
    4       3.9    13      35.5
    5       4.5    17      34.5
    6       5.0    20      37.0
    7       5.2    ...     ...
    8       5.7
    9       6.8
    10      7.9
    ...     ...

Result:

    index1  data1  data2
    1       2.1    NaN
    2       3.3    30.5
    3       3.5    30.5
    4       3.9    30.5
    5       4.5    30.5
    6       5.0    32.0
    7       5.2    32.0
    8       5.7    32.0
    9       6.8    35.0
    10      7.9    35.0
    ...     ...    ...

EDIT:
I think the following post is close to what I need, but I’m not sure how to transform the solution to fit my problem.
http://www.mathworks.com/matlabcentral/newsreader/view_thread/260139

EDIT (Several Months Later):
I’ve recently come across this excellent little function that I think may be of use to anyone who lands on this post:

function yi = interpLast(x,y,xi)
%INTERPLAST Interpolates the input data to the last known value.
% Note the index data should be input in ASCENDING order.
inds = arrayfun(@findinds, xi);
yi = y(inds);

function ind = findinds(val)
    ind = find(x<=val,1,'last');
    if isempty(ind)
        ind = 1;
    end
end

end

Credit goes here: http://www.mathworks.com/support/solutions/en/data/1-48KETY/index.html?product=SL&solution=1-48KETY

Answer 1

The problem is one of run length decoding. See section 15.5.2 of Matlab array manipulation tips and tricks (which is an eye-opening read for any Matlab enthusiast).

Here’s using the method with your example (I’m using octave but the code is identical for Matlab):

octave:33> a=[2,30.5;6,32;9,35;13,35.5;17,34.5;20,37]
a =

    2.0000   30.5000
    6.0000   32.0000
    9.0000   35.0000
   13.0000   35.5000
   17.0000   34.5000
   20.0000   37.0000

octave:34> i=a(:,1)-1
i =

    1
    5
    8
   12
   16
   19

octave:35> j=zeros(1,i(end))
j =

   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0

octave:36> j(i(1:end-1)+1)=1
j =

   0   1   0   0   0   1   0   0   1   0   0   0   1   0   0   0   1   0   0

octave:37> j(1)=1
j =

   1   1   0   0   0   1   0   0   1   0   0   0   1   0   0   0   1   0   0

octave:38> val=a(:,2)
val =

   30.500
   32.000
   35.000
   35.500
   34.500
   37.000

octave:39> x=val(cumsum(j))
x =

   30.500
   32.000
   32.000
   32.000
   32.000
   35.000
   35.000
   35.000
   35.500
   35.500
   35.500
   35.500
   34.500
   34.500
   34.500
   34.500
   37.000
   37.000
   37.000

And pad the beginning with NaN as needed.