how to do interpolation?

Hi all,

I worked with the time series of the sea level data (non-tidal residual. thus, there is no periodicity in the time series. could you please help me to do interpolation because it contains some NaN values.

Thank you in advance.

Answer 1

[Kelly Kearney]

A little more explanation of what you're trying to do is needed here. Are you trying to interpolate from scattered data to a regular grid? If so, your best option is probably to drop the NANs prior to interpolation. Or do you want to fill in some of the missing spots in an already-gridded dataset? Or something else?

[John BG]

I was about to write something, but found

http://uk.mathworks.com/matlabcentral/fileexchange/4551-inpaint-nans?s_tid=srchtitle

does it work for your question? if it doesn't let know to see what can else be done.

[John D'Errico]

Yes, inpaint_nans does work on 1-d data. So, given a vector of values as a time series, some of which are NaNs, it will interpolate the NaNs. It will also extrapolate if there are NaNs at one or both ends.

[Lilya]

Kelly, yes. I have to fill the missing values in the array.

[Lilya]

Johns', It does not work. I tried it already.

[Kelly Kearney]

Again, can you give a little more detail? Just saying "it doesn't work" doesn't help us, since inpaint_nans should be exactly what you need:

% Some fake data with missing points

x = 1:100;
y = sin(x/pi);
y(rand(100,1)> 0.8) = NaN;

% Fill the NaNs

y2 = inpaint_nans(y);

% Plot

plot(x, y, '-x', x, y2, ':')

[Lilya]

THANKX a lot.