Replace NaN with median per column

Hi,

I have a matrix where I have NaN values in a few columns.

I would like to fill the NaN values with the median value of that column. How do I do that?

Answer 1

[KSSV]

Read about fillmissing.

Or USe:

A = rand(20,1) ;   % data for demo 
A(randperm(20,5)) = NaN ; % insert nans 
M = nanmedian(A) ; % get median
A(isnan(A)) = M % replace nans with medians

Answer 2

[Wan Ji]

It is convenient to use fillmissing function to get what you want

matOut = fillmissing(matrixIn, 'linear', 'EndValues','nearest')

[Konvictus177]

What if an entrie column is NaN?

[Wan Ji]

if a whole NaN column exists, you can transpose the matrix and use the above method. btw, meaningless if a column is all NaN, isn't it?

[Yazan]

This does linear interpolation of neighboring non-nan values. It does not replace nan values with the median.

[Wan Ji]

@Konvictus177

Answer 3

[Stephen]

M = randi(3,5,7);
M(randi(numel(M),1,9)) = NaN

M = 5×7

1 NaN 1 1 2 3 1 NaN 1 2 NaN 1 1 3 NaN 3 NaN 1 2 1 NaN 2 NaN 3 2 3 2 2 1 1 2 NaN 1 3 2

V = median(M,1,'omitnan')

V = 1×7

1 1 2 1 2 2 2

X = isnan(M);
M(X) = repelem(V,sum(X,1))

M = 5×7

1 1 1 1 2 3 1 1 1 2 1 1 1 3 1 3 2 1 2 1 2 2 1 3 2 3 2 2 1 1 2 1 1 3 2

Answer 4

[Yazan]

clc, clear
x1 = randn(5, 5);
x1(randi(numel(x1), 1, 5)) = nan;
x2 = fillmissing(x1, 'movmedian', size(x1,1)*2, 1);

display(x1)

x1 = 5×5

NaN 0.1826 NaN 0.3875 -0.7928 1.3983 -0.1527 0.1261 0.1482 -0.8491 1.4623 0.7005 -0.0511 -0.1259 1.1562 0.8948 0.0311 0.0064 NaN NaN -0.2615 -0.1371 -1.7245 0.9723 0.5918

display(x2)

x2 = 5×5

1.1466 0.1826 -0.0223 0.3875 -0.7928 1.3983 -0.1527 0.1261 0.1482 -0.8491 1.4623 0.7005 -0.0511 -0.1259 1.1562 0.8948 0.0311 0.0064 0.2678 -0.1005 -0.2615 -0.1371 -1.7245 0.9723 0.5918