Why does resample lose values when applied to vector with NaN's?

I am resampling a vector with a few final values equal to NaN. The output of resample.m is shorter than expected. Could somebody explain why this happens? According to MATLAB manual , NaN's should be safely ignored.

Here's the code that reproduces the problem:

test = [randn(80,1); nan(20,1)];
resampled = resample(test,1,3);
figure, plot(test,'b')
hold on, plot(1:3:100,resampled,'r')
downsampled = downsample(test,3);
hold on, plot(1:3:100,downsampled,'k')

and the figure:

Note that "downsample" doesn't have this problem.

resample.m and firls.m are the ones provided by MATLAB:

which resample
C:\Program Files\MATLAB\R2014b\toolbox\signal\signal\resample.m
which firls
C:\Program Files\MATLAB\R2014b\toolbox\signal\signal\firls.m

Thank you!

Answer 1

[Akira Agata]

The function downsample decreases the sampling rate by extracting every n-th sample. On the other hand, resample applies an antialiasing FIR filter when changing the sampling rate. That's the reason why outputs of these function differ each other.