Error in the specification of the window when using movmad

Hello,

I got a matrix called "A" [nxm], where n is the number of observations and m is the number of variables.

I want to perform a movmean and movmedian, but instead of use the size of the window kb and kf the same for all the variables, i created a matrix B [1xm], with the values of "k" to use for each variable of the matrix table

the code to do this should be?

      n = 969120;
      m = 41;
      result1 = NaN(n,m);
      result2 = NaN(n,m);
      for j = 1:m
      for j = 1:n
      result1(i,j) = movmad(A(i,j),[B(1,j) B(1,j)],'omitnan');
      result2(i,j) = movmedian(A(i,j),[B(1,j) B(1,j)],'omitnan');
      end
      end

because when i run this in a script, i get the following error:

"error using movmad, window length must be a finite positive scalar or 2-element vector of finite nonnegative scalars"

[Walter Roberson]

I recommend against using table as the name of a variable, as you are going to confuse readers who know that table() creates table data objects.

table(i,j) is going to be either a scalar numeric value or a scalar struct or a scalar cell array or a single row and column of a table data object (we do not know which because we did not see the initialization.) The only one of those that can have movmad() applied to it is the scalar numeric value, and movmad() is useless on scalar numeric values.

I speculate that your A values include non-integers or 0s or nan.

[Tiago Dias]

table and A are matrix, are just numbers, it is just a name ok.

My "table" has both numbers and nan, thats why i use omitnan. It was working for just the same value of A for all variables, now that i use severall, it doesnt work. i will upload my data so u can take a look

[Walter Roberson]

"it is just a name ok."

If you expect other people to read your code (such as if you are asking them to help debug your code) then it is highly recommended that you use names that are easy to understand, avoiding using variable names that are open to interpretation or which require close reading of all of the lines. Avoiding, in other words, using the names of any of the common functions as a variable name.

When you make code difficult for people to read, they might not bother to read it.

[Tiago Dias]

ok i will change the name of the variables to be easier to read and understand, sorry

[Tiago Dias]

I think now it is more understandable

[Walter Roberson]

You still have the difficulty that you appear to be apply movmad() and movmedian() to scalar numerics.

[Tiago Dias]

i don't understand that, could you use an example?

Answer 1

[Walter Roberson]

      if any(isnan(B)) || any(B(<0)) || any(B~=floor(B))
        error('B must be non-negative integers');
      end

      n = 969120;
      m = 41;
      result1 = NaN(n,m);
      result2 = NaN(n,m);
      for i = 1:n
        result1(i,:) = movmad(A(i,:), [B(i) B(i)], 'omitnan');
        result2(i,:) = movmedian(A(i,:), [B(i) B(i)], 'omitnan');
      end

[Tiago Dias]

Perhaps i am mistaken, but what i want to do is to applot movmad for the 1st column of A with the 1st value of B, then movmad to the 2nd column of A the 2nd value of B.

and what o wrote applies in rows no?

[Walter Roberson]

      if any(isnan(B)) || any(B(<0)) || any(B~=floor(B))
        error('B must be non-negative integers');
      end

      n = 969120;
      m = 41;
      result1 = NaN(n,m);
      result2 = NaN(n,m);
      for i = 1:m
        result1(:,i) = movmad(A(:,i), [B(i) B(i)], 'omitnan');
        result2(:,i) = movmedian(A(:,i), [B(i) B(i)], 'omitnan');
      end