calculating rmse between observations and estimates of different size

I would like to calculate rmse between observations and estimates of different size. This are two timeseries, one 84084x1 and the second is 315360x1.

Answer 1

[Thomas]

Your data sets (data and estimate) have different sizes. Do you have NaN's in your estimates or are you taking multiple estimates/observations?

You might want to equalize the size of the dataset (by removing the unwanted observations) and then finding the rmse

 rmse=sqrt(sum((data(:)-estimate(:)).^2)/numel(data));

[Sergio]

The are gaps in the observation. Is there a way to fill those gaps in a time step with a zero or interpolation?

[Thomas]

You could try interp1 (nearest neighbor interpolation) or something similar http://www.mathworks.com/help/matlab/ref/interp1.html

[Thomas]

I just realized that you have a time series and if you have the finance toolbox you can use the fillts command http://www.mathworks.com/help/finance/fillts.html

Answer 2

[Youssef Khmou]

You can padd the small vector with zero or Interpolate as mentioned by @Thomas :

 % Given your vectors r1 r2
 r1=randn(315360,1);
 r2=randn(84084,1);
 N1=length(r1);
 N2=length(r2);
 Ratio=floor(N1/N2);
 r22=interp(r2,Ratio);
 Diff=N1-length(r22);
 r22=[r22;zeros(Diff,1)];
 plot(r1), hold on, plot(r22,'r'),
 RMSE=sqrt(mean(((r22-r1).^2)))

You can also use the functions downsample or upsample

[Sergio]

Hi Youssef,

I tried your code but it gives me error:

Error using horzcat Dimensions of matrices being concatenated are not consistent.

Error in usgs (line 107) r22=[r22 zeros(1,Diff)];

Thanks though. It has potential

[Youssef Khmou]

Sergio, transpose vectors r1=r1'; r2=r2' or try again the edited code .