Removing all non NaN Values to a new matrix.

Hi,

I have got a 261x3733 double array temp with numbers and NaN.

I would like to get in matrix a only the nonnan values. Conditions: a should be of the same size as temp and all nonnan values shouldn't change place.

I tried this code:

a=temp(~isnan(temp))

However I get the vector with size 330227x1

Do you have any suggestions for solving this issue?

[Adam]

What you want in place of the NaNs? You can't have nothingness is elements of a numeric array. If you extract only the non-nan values then you inevitably loose the shape of the matrix.

[BN]

I'm Agree with Adam.

So you want to remove NaN and save the result in a new matrix BUT you want a new matrix to have exactly the same size as an original matrix?

What do you want to instead of NaN?

This data seems to be the temperature. You can fill them using the nearest station linear regression or using fillmissing function.

Answer 1

[JESUS DAVID ARIZA ROYETH]

a solution:

a=sparse(temp)
a(isnan(a))=[]

Answer 2

[BdS]

I would like to apply the function winsorising on the elements with values only. The function winsorising accounts unfortunatelly for NaN elements.

[Adam]

I don't know what the function is, but can you not just pass the non-nan values to it and then if you want the result in your matrix you can put the results back in to the non-nan locations of a new result matrix.

Answer 3

[BdS]

This is the function I would like to use

https://www.mathworks.com/matlabcentral/fileexchange/32327-winsorising-data