indexing with isnan in multidimensional arrays

Question

Sagar Parajuli 2021-8-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/896867-indexing-with-isnan-in-multidimensional-arrays

评论： Walter Roberson 2021-8-13

采纳的回答： Walter Roberson

lon_nonan_reshape_use.mat

在 MATLAB Online 中打开

I have this code:

load lon_nonan_reshape_use;
idx = ~isnan(lon_nonan_reshape_use);
lon_nonan_adj = lon_nonan_reshape_use(idx);

In the above example, lon_nonan_reshape_use and idx both are of size n*2 so I expect lon_nonan_adj also to be of n*2. MATLAB gives the results in n*1 dimension because lon_nonan_adj(:, 1) is not equal to lon_nonan_adj(:, 2) which is sensible. But I still want to get lon_nonan_adj as a n*2 matrix with NaNs retained in the place where there were NaNs previously. Could you please help?

4 个评论
显示 2更早的评论隐藏 2更早的评论

DGM 2021-8-12

"But I still want to get lon_nonan_adj as a n*2 matrix with NaNs retained in the place where there were NaNs previously."

If you take matrix A, remove all the NaNs and then fill all the empty spots left over with NaN such that the size is preserved, then you're back at A. I don't see how the operation isn't superfluous.

Sagar Parajuli 2021-8-13

I think you are right, so it looks like I should get rid of all NaNs and deal with the no-nan data only as answered below. Thank you.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2021-8-12

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/896867-indexing-with-isnan-in-multidimensional-arrays#answer_765502

在 MATLAB Online 中打开

load lon_nonan_reshape_use;
idx = ~all(isnan(lon_nonan_reshape_use),2);
lon_nonan_adj = lon_nonan_reshape_use(idx,:);

This retains rows provided there is at least one non-nan in the row.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Walter Roberson 2021-8-13

在 MATLAB Online 中打开

If you want to remove all rows with NaN anywhere in the row, then

rmmissing(lon_nonan_reshape_use)

请先登录，再进行评论。

Answer 2

Chunru 2021-8-12

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/896867-indexing-with-isnan-in-multidimensional-arrays#answer_765507

在 MATLAB Online 中打开

% A small matrix with nans
a=randn(6, 3);
a([2 11 13])=nan
a = 6×3
   -0.5340    0.4891       NaN
       NaN    0.2906    0.4462
   -1.2660   -1.1701    2.0372
   -0.3616   -0.4775    0.3251
    0.6966       NaN    0.1013
    0.4787    1.2229   -0.6159
% idx
idx = ~isnan(a)
idx = 6×3 logical array
   1   1   0
   0   1   1
   1   1   1
   1   1   1
   1   0   1
   1   1   1
a(idx)
ans = 15×1
   -0.5340
   -1.2660
   -0.3616
    0.6966
    0.4787
    0.4891
    0.2906
   -1.1701
   -0.4775
    1.2229

So a(idx) is a colum matrix based on the original a by removing all nans and arranged in a column vector.

If you want to remove all rows with nans (it looks so for your problem), then you can do the following

idx1 = all(~isnan(a), 2)
idx1 = 6×1 logical array
   0
   0
   1
   1
   0
   1
a(idx1, :)
ans = 3×3
   -1.2660   -1.1701    2.0372
   -0.3616   -0.4775    0.3251
    0.4787    1.2229   -0.6159

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Sagar Parajuli 2021-8-13

I like your last solution because ultimately that is what I need (data with all NaNs removed), to input to 'griddata'. Thank you for forseeing my problem.

请先登录，再进行评论。