Deleting missing values (different amount and distribution of NaNs for each column)

Question

Peter P 2017-12-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/373893-deleting-missing-values-different-amount-and-distribution-of-nans-for-each-column

评论： Jos (10584) 2017-12-20

I have two matrices of the same size (17x82). Matrix A has no missing values. Matrix B has a different amount and distribution of NaNs for each column. I want to delete all missing values from matrix B as well as the values of matrix A that correspond to the NaNs of matrix B.

I tried: A_Without_NaN = A(~isnan(B)); However, reshaping does not work because the number of NaNs in matrix B varies for each column.

Any suggestions? Thanks in advance, Peter

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jos (10584) 2017-12-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/373893-deleting-missing-values-different-amount-and-distribution-of-nans-for-each-column#answer_297129

在 MATLAB Online 中打开

You cannot concatenate different length vectors into a single matrix. You could store each column of A into a separate cell

A_wn = arrayfun(@(k) A(~isnan(B(:,k)),k),1:size(A,2),'un',0)
% A_wn is a cell array. The k-th cell, A_wn{k}, holds the
%    'relevant' values of the k-th column of A

However, it might be preferred to leave the nans and use nan-specific functions like nansum and nanmean.

2 个评论
显示无隐藏无

Steven Lord 2017-12-20

If you're using release R2015a or later, you can call sum and mean with the 'omitnan' parameter instead of calling nansum or nanmean. See the second item in the Mathematics section of the Release Notes for R2015a.