3

This is a small thing but has been bothering me for a while now, so I thought I would let the crowd solving begin :)

I have a matrix with timestamps and a corresponding logical value (or 1/0), i.e.

of = [-inf 0 10 15 190 inf; 1 0 0 1 1 0]'

and an another time vector, e.g.

t = 0:0.1:1e3;

or whatever, you get the point :)

Now how do I (read: would you) inter-/extrapolate the logical infomation in of so it matches the timestamps in t, but with the interpolated logicals always assuming the last or current value, not a future one?

Don't know if that makes sense, but here is the expected output given of and t2

t2 = [0 5 14 16]
output = [0 0 10 15; 0 0 0 1]'

where the first column of output is the time of of used in interpolation. If I use interp1 and the 'nearest' algorithm, it will give

interp1(of(:,1), of, t2, 'nearest')
output = [0 10 15 15; 0 0 1 1]'

which is not exactly what I want.

Juhl
  • 473
  • 4
  • 17

1 Answers1

4

Assuming your vectors are sorted you could try that, which seems to work with your example although I did not tested it extensively:

of=[-inf 0 10 15 190 inf; 1 0 0 1 1 0]';
t2 = [0 5 14 16];
index=floor(interp1(of(:,1),(1:size(of,1))',t2'));
output=of(index,:);

Hope it helps.

The default method used by interp1 is linear, which works best with your condition because you do not want the "nearest" neighbor but the first lower or equal neighbor (as far as I understand this). Therefore a simple truncation of the interpolated timestamps index gives you the result.

Aabaz
  • 3,106
  • 2
  • 21
  • 26
  • F*** me thats the stuff! :D Some times you can just get lost in complexity, I guess that is what happend here :) Also, infs can be ignored if one rethinks the problem, so this will work. – Juhl Sep 05 '11 at 14:10