How to remove spikes without modifying the dataset?

Hi,

I need to remove spikes from a signal, but I don't want to modify the dataset. Using the medfilt1 function, it can remove spikes but it also interpolates all the dataset. As an example, given the following dataset

data = [1,2,10,1,3,2,1,2];

the function gives

>> medfilt1(data)

ans =

     1 2 2   3 2 2 2 1

but what I want is

     1 2 NaN 1 3 2 1 2

i.e. I just want to remove the spike (10) replacing it with a NaN.

Is this possible in some way?

Thank you

Answer 1

[Star Strider]

Use the isoutlier function:

data = [1,2,10,1,3,2,1,2];

TF = isoutlier(data);                               % Function Introduced In R2017a
data(TF) = NaN

producing:

data =
     1     2   NaN     1     3     2     1     2

To keep the original ‘data’ vector, first make a copy of it, then do the analyses.