Remove specific outliers from double

I want to remove outliers from a double using the rmoutliers function, combied with "datavariables" to select the columns from which the outliers should be removed. See the code below.

mData3=rmoutliers(mData3,'percentiles',[5 95],'DataVariables',[3 4]);

The addition of the "datavariables" argument messes with the function and leads to no output. It seems that it only works with a table datatype, however I am unable to convert the double to a table. (I also need the data to be in the double-form)

Is there another solution?

Kind regards

[Dyuman Joshi]

"however I am unable to convert the double to a table."

Why is that?

Also, after converting the data into a table, and removing the outliers for 3rd and 4th column (if you there are any) you will not be able to convert the data back into a double array, because the size of the output will not match for concatenation.

[Konstantin Koerber]

Hey, thank you for your answer. I tried array2table but it didn't work. Since I want to keep the data in double array anyways, I'd prefer a different solution to only remove the outliers in the columns I want to. Any ideas?

[Dyuman Joshi]

"I tried array2table but it didn't work."

Can you attach your code and data?

Suppose your data is this -

A = magic(12);
A = A(:,1:4);
A(3,3) = -200;
A(4,4) = -300;
disp(A)

   144     2     3   141
    13   131   130    16
    25   119  -200    28
   108    38    39  -300
    96    50    51    93
    61    83    82    64
    73    71    70    76
    60    86    87    57
    48    98    99    45
   109    35    34   112
   121    23    22   124
    12   134   135     9

%3rd column
isoutlier(A(:,3),"percentiles",[5 95])'

ans = 1×12 logical array
   0   0   1   0   0   0   0   0   0   0   0   1

%4th column
isoutlier(A(:,4),"percentiles",[5 95])'

ans = 1×12 logical array
   1   0   0   1   0   0   0   0   0   0   0   0

What should be the final output here?

Answer 1

[Steven Lord]

Use normal indexing to replace the outliers in certain columns using filloutliers.

A = magic(5);
A(3, :) = A(3, :) + 100 % Create outliers

A = 5×5

17 24 1 8 15 23 5 7 14 16 104 106 113 120 122 10 12 19 21 3 11 18 25 2 9

A(:, [2 4]) = filloutliers(A(:, [2 4]), NaN)

A = 5×5

17 24 1 8 15 23 5 7 14 16 104 NaN 113 NaN 122 10 12 19 21 3 11 18 25 2 9

Note that you can't remove outliers in this scenario, as that would lead to a matrix with different length columns and that is not allowed in MATLAB numeric arrays.

A(:, [1 5]) = rmoutliers(A(:, [1 5]))

Unable to perform assignment because the size of the left side is 5-by-2 and the size of the right side is 4-by-2.