Converting/resampling 10-min interval timseries data to 15-min

Hello

I have time series data taken at 10-min intervals (see below). I would like to somehow obtain values at 15-min intervals from this data.

(% Y M D H MI S VAL)

2008  1  1  0  0  0  1.68400000000000
2008  1  1  0  10  0  1.71400000000000
2008  1  1  0  20  0  1.74400000000000
2008  1  1  0  30  0  1.77300000000000
2008  1  1  0  40  0  1.80300000000000
2008  1  1  0  50  0  1.83400000000000
2008  1  1  1  0  0  1.86500000000000
2008  1  1  1  10  0  1.89800000000000
2008  1  1  1  20  0  1.93400000000000
2008  1  1  1  30  0  1.97200000000000
2008  1  1  1  40  0  2.01200000000000
2008  1  1  1  50  0  2.05500000000000

I have tried using the resample function as follows:

ts2 = timeseries(val);
val_int = resample(ts2, 0:15/24/60:718); %%(There are 719 values in the original timeseries)

...This really doesn't work very well!

Is resample the right tool to use? If not, could somebody please give any hints? I was thinking of maybe some sort of moving average?

Many thanks in advance

Ozgun

Answer 1

[Star Strider]

I’m not clear on what you are doing. You seem to be creating a timeseries object using only one of your variables, VAL. I do not have much recent experience with timeseries objects, so I did the interpolation on your data (that I call Mtx) with interp1:

Mtx = [2008  1  1  0  0  0  1.68400000000000
2008  1  1  0  10  0  1.71400000000000
2008  1  1  0  20  0  1.74400000000000
2008  1  1  0  30  0  1.77300000000000
2008  1  1  0  40  0  1.80300000000000
2008  1  1  0  50  0  1.83400000000000
2008  1  1  1  0  0  1.86500000000000
2008  1  1  1  10  0  1.89800000000000
2008  1  1  1  20  0  1.93400000000000
2008  1  1  1  30  0  1.97200000000000
2008  1  1  1  40  0  2.01200000000000
2008  1  1  1  50  0  2.05500000000000];

Mtx1 = [datenum(Mtx(:,1:6))  Mtx(:,7)];
Min15 = 1/(24*4);   % Four 15-min segments/hour; here, fractions of a day
TIntrpVal = [Mtx1(1,1):Min15:Mtx1(end,1)];
Mtx2 = interp1(Mtx1(:,1),Mtx1,TIntrpVal);
TestMtx2 = [datevec(Mtx2(:,1) ) Mtx2(:,2)]  % Test to be sure it works

produces:

TestMtx2 =  
    2008   1   1   0   0   0  1.684
    2008   1   1   0  15   0  1.729
    2008   1   1   0  30   0  1.773
    2008   1   1   0  45   0  1.819
    2008   1   1   1   0   0  1.865
    2008   1   1   1  15   0  1.916
    2008   1   1   1  30   0  1.972
    2008   1   1   1  45   0  2.033

Does that do what you want?

[Ozgun]

Hello

Thanks for the reply. Yes that looks to be exactly what I wanted.

I did come across interp1 - should have looked at it more closely!

Many thanks for the help.

Ozgun

[Star Strider]

My pleasure!

I’m still learning things about MATLAB functions and all they can do. It seems to be a never-ending process.

Creating and using a timeseries object is certainly an option. You have to create it differently than you did for it to do what you want.

[Stefan Kermer]

Hi there,

it is astonishing, that with 5 lines of code you produce such a nice result. Nevertheless it produces wrong results.

Can someone help me to make little adaptions of that code to get it right?

... As the original data is in 10 minutes resolution and the target resolution is a 15 minutes interval it is not correct to interpolate linear...

A correct interpolation function would need to implement a weight with 2/3 first 10 minutes and 1/3 second 10 minutes for the first 15 minutes.. and so forth Do you know what I mean?

Stefan

[Stefan Kermer]

Hi guys,

found the right code...

Matlab is amazing

W_10 = (1:52560)'; % Dummy-Daten

n_10 = length(W_10);

n_15 = n_10 * 2 / 3;

W_15(1:2:n_15) = (W_10(1:3:n_10) * 2 + W_10(2:3:n_10)) / 3;

W_15(2:2:n_15) = (W_10(2:3:n_10) + W_10(3:3:n_10) * 2) / 3;

http://www.gomatlab.de/10-minuten-in-15-minuten-werte-wind-etc-umrechnen-t19807.html