15-Moonstone

Solved

filterNaN

Forum|Forum|1 day ago
July 29, 2026
2 replies
49 views

Hi There ,

I want to use this function ( filterNAN) , but I’m not sure how it works . I have a nested array f2 . I tried three different approaches but non is working , in f2 I’m using M and sigma they also have NaN in their values , but not all the values are NaN.

Regards,

Yusra

Best answer by LucMeekes

There’s a couple of functions involving NaNs.

First off there is IsNaN(x), which tells you if x is a NaN (or not):

Realise that IsNaN is the only way to test if an item is a NaN. You cannot use comparative expressions to test for NaNs, as shown here:

Defines n to be a NaN

But the comparison of n with NaN results in 0 (False). That is to say, n does not equal NaN, even though it is defined NaN.

Even more clear:

NaN is unequal to NaN. Again: you cannot compare with NaNs.

Then there is markNaN(M,I), wich takes a Matrix and a vector of row/column pairs and marks the items in M indicated by pairs in I as a NaN:

If

and

then

(You can also do this with a vector:

Set the two items of the (first) vector with the indexes of the second vector to NaN.

)

You can find I back from MN with matchNaN(x), which finds the r/c pairs in x that contain a NaN:

(On the vector:

)

Finally there's filterNaN(x), that filters out all rows of x that contain a NaN. So with no NaNs it gives:

But on the modified matrix that contains NaNs it gives:

just the last row of MN, as the first two rows each contain a NaN.

Now note that for an item to be a NaN, it must be a scalar (single value).

Tells you that the vector [1 NaN] is NOT a NaN (because it is not a scalar)

And so:

filterNaN returns the full vector (of vectors) because it contains no NaN scalars

but

Has filterNaN working with only the first element of the nested vector, that element is a vector and contains a NaN in the second row. Hence it returns a vector with just the first element of it.

Success!
Luc

Answer

23-Emerald IV

There’s a couple of functions involving NaNs.

First off there is IsNaN(x), which tells you if x is a NaN (or not):

Realise that IsNaN is the only way to test if an item is a NaN. You cannot use comparative expressions to test for NaNs, as shown here:

Defines n to be a NaN

But the comparison of n with NaN results in 0 (False). That is to say, n does not equal NaN, even though it is defined NaN.

Even more clear:

NaN is unequal to NaN. Again: you cannot compare with NaNs.

Then there is markNaN(M,I), wich takes a Matrix and a vector of row/column pairs and marks the items in M indicated by pairs in I as a NaN:

If

and

then

(You can also do this with a vector:

Set the two items of the (first) vector with the indexes of the second vector to NaN.

)

You can find I back from MN with matchNaN(x), which finds the r/c pairs in x that contain a NaN:

(On the vector:

)

Finally there's filterNaN(x), that filters out all rows of x that contain a NaN. So with no NaNs it gives:

But on the modified matrix that contains NaNs it gives:

just the last row of MN, as the first two rows each contain a NaN.

Now note that for an item to be a NaN, it must be a scalar (single value).

Tells you that the vector [1 NaN] is NOT a NaN (because it is not a scalar)

And so:

filterNaN returns the full vector (of vectors) because it contains no NaN scalars

but

Has filterNaN working with only the first element of the nested vector, that element is a vector and contains a NaN in the second row. Hence it returns a vector with just the first element of it.

Success!
Luc

Author

15-Moonstone

thank you so much I appreciate your help