Removing NaN from matrix

Hello everybody,

I have a 61x9 double matrix. The columns look like this:

Column 1: 0 1 2 3 4 5 6 7 ... NaN NaN NaN

Column 2: 0 2 3 4 5 6 7 8 4 5 6 3 NaN NaN

Column 3: 0 5 6 6 3 3 6 7 8 5 3 3 6 7 8 5 . . .

That means, every column has a certain amount of values and is filled up with NaNs to get a rectangular double matrix all in all.

Now I want to remove all the NaNs. And I am not quite sure if this is possible, because after removing the NaNs, the size of each column would differ.

For now I have tried something like this:

A(find(isnan(A)))=[]

This code removes the NaNs, but the result is a 1x445 double, where all columns are just add consecutively.

I hope you guys can help me!

Cheers

Christian

[Ganesh Hegade]

Its better you should replace the NaN with some values.

    A(isnan(A)) = -1;

[Pampa Dey]

how to remove nans from a column?

I have multiple .txt files in which some of them have completely nan values within my interested column how can we remove them in matlab?

could anyone please help?

Answer 1

[Jan]

Exactly: This cannot work, because all columns of a matrix must have the same number of rows. You cannot remove all NaNs and keep the shape of the matrix. You have to decide, what you want instead:

X = X(~any(isnan(X, 2)))
X = X(~all(isnan(X, 2)))
X = X(~isnan(X))

Or perhaps you want a cell array containing the different columns vectors?

[Christian]

Ok, that's what I thought. So I will continue using cells instead.

[Ashishkumar Gupta]

did not work for a .struct!!!

Any other way out to remove NaNs from a struct having 50-60 variables

Answer 2

[Biswajit Ojha]

"X = X(~any(isnan(X, 2)))

X = X(~all(isnan(X, 2)))

X = X(~isnan(X)) "

dear Jan,

can you please explain why you have used those '2' s in the command?

[Yuanhanqing Huang]

I think there are some little errors. Maybe the right code should be:

X = X(~all(isnan(X), 2))

It means that we find the rows the elements of which are all nan. '2' denotes the results are in column. That being said, the any operation is implemented regarding each row.