I have arrays like this:
[1 NaN NaN]
[NaN 2 NaN]
[NaN NaN 3]
How can I merge them to
[1 2 3]
?
-
You have N arrays of length N with a single number at different positions? Or is it a matrix (sorted or unsorted)?maraca– maraca2016-08-02 23:24:32 +00:00Commented Aug 2, 2016 at 23:24
-
There are actually N matrices where each row is either filled with NaN or filled with numbers. Unsorted.stollenm– stollenm2016-08-02 23:27:32 +00:00Commented Aug 2, 2016 at 23:27
-
And when more than one array has a non-Nan number how do we merge them? Or will there only ever by one non-Nan number per index?Jerry Jeremiah– Jerry Jeremiah2016-08-03 00:51:10 +00:00Commented Aug 3, 2016 at 0:51
Add a comment
|
4 Answers
Concatenate the vectors into a matrix and then compute the mean of each column discarding NaN's:
vectors = {[1 NaN NaN], [NaN 2 NaN], [NaN NaN 3]};
result = mean(vertcat(vectors{:}), 1, 'omitnan'); % or nanmean(vertcat(vectors{:}),1)
1 Comment
stollenm
Thanks, that does exactly what I was looking for. I think its simpler like this though:
vectors = [[1 NaN NaN]; [NaN 2 NaN]; [NaN NaN 3]]; result = mean(vectors,'omitnan')You can concatenate all the matrices into a single, big matrix and then use logical indexing to take the numbers out:
A = [1 NaN NaN 1];
B = [NaN 2 1 NaN];
C = [3 NaN 2 3];
D = [A B C];
E = D(~isnan(D))
Now E looks like this:
E =
1 1 2 1 3 2 3
Comments
You can replace the NaNs with zeros:
x = [1 NaN NaN]
y = [NaN 2 NaN]
z = [NaN NaN 3]
x(isnan(x)) = 0
y(isnan(y)) = 0
z(isnan(z)) = 0
x + y + z
Comments
You can combine all your matrices like so (this puts A on top of B):
C = [A ; B];
Then you can detect the cells that are not NaN:
D = ~isnan(C);
After that you can select with those logicals:
e = C(D);
This is now a vertical vector, you can make it horizontal with:
e = e';
