# how to remove NAN from a double inside a cell array?

53 views (last 30 days)
Farshid Daryabor on 23 Jan 2020 at 14:31
Commented: Farshid Daryabor on 23 Jan 2020 at 17:15
can some one how can I remove NAN from a double inside a cell array (attached file)?

Image Analyst on 23 Jan 2020 at 15:04
That's not an answer. Matrices have to be rectangular - they can't have ragged edges. So the NaN's must be replaced by something, not removed. So what do you want to replace them with?
Guillaume on 23 Jan 2020 at 16:15
You haven't answered my question and we can't really answer yours until you do. You may get some answers that do something but possibly not what you want.
Given the matrix M in my example, you could
• Delete rows that have any NaN, in my example this would also remove the numbers 7, 8, 9 and 10, ending up with:
M = [1 2 3
4 5 6]; %all the other rows have NaN
• Delete rows that are all NaN, this still leaves some NaN, but doesn't delete any number, ending up with:
M = [1 2 3
4 5 6
7 8 NaN
9 10 NaN
11 NaN NaN];
• Delete columns that have any NaN, in my example this delete all columns, ending up with
M = []
• Delete columns that are all NaN. In my example this doesn't delete any rows. You get the same M
• Delete all the NaNs, since you can't have holes in a matrix, you'd end up with a vector:
M = [1 4 7 9 11 2 5 8 10 3 6] %in this order
• do something else that you haven't explained.
Farshid Daryabor on 23 Jan 2020 at 17:15
Each columns are corresponds to a latitude and longitude to introduce seawater temperature profiles. I want to keep the matrix structure as is, but delete the NaNs from each column.
I tried to interpolate NaNs, but problem is converting cell to number, I used the following function, encountering with error, Out = cellfun(@cell2mat, (inpaint_nans(mycelldata,0)) ); Any comment to improve.

KSSV on 23 Jan 2020 at 14:55
You can fill the nans using fillmissing. Also you can do interpolation ang the values at the places of nan. Read about interp1.

Farshid Daryabor on 23 Jan 2020 at 16:02
Yes
KSSV on 23 Jan 2020 at 16:41
If A is your matrix with Nan's.
iwant = A(~isnan(sum(A,2)),:)
Guillaume on 23 Jan 2020 at 16:48
Note that this is a solution to my second bullet point.
A better way (probably faster) to do the above is:
newA = A(~any(isnan(A), 2), :)