movavg with custom type and weights

I have difficulties using the movavg with custom type and weight. The minimum reproducible code is given below

It complained that the weight vector is not an array with all values <= 1 but isn't B clearly fulfil this criteria? Or am I missing something?

[Jan]

Please post code as formatted text, not as screenshot. Then it is much easier to re-use it by copy&paste.

Answer 1

[Bruno Luong]

movavg requires first array is a column vector or matrix

A = 1:5;
B = [0.5 0.5];
movavg(A(:), "custom", B)

ans = 5×1

1.0000 1.5000 2.5000 3.5000 4.5000

% or reshape back
reshape(movavg(A(:), "custom", B), size(A))

ans = 1×5

1.0000 1.5000 2.5000 3.5000 4.5000

movavg(A, "custom", B) % error

Error using movavg
Expected weights vector to be an array with all of the values <= 1.

The error message is for sure misleading.

Furthermore according to the doc

"To compute moving average with custom weights, the weights (w) are first normalized such that they sum to one:

W(i) = w(i)/sum(w), for i = 1,2,...,N"

Not serious TMW on document-error handling this function.

Answer 2

[Image Analyst]

I never heard of movavg - I guess it's only in the Financial Toolbox. You can use conv or movmean instead:

A = [1,2,3,4,5]; % Data to be filtered
B = [0.5, 0.5]; % Weights
C = conv(A, B, 'valid')

C = 1×4

1.5000 2.5000 3.5000 4.5000

% or use movmean
C = movmean(A, 2)

C = 1×5

1.0000 1.5000 2.5000 3.5000 4.5000

Depending on what you want your output to be (how you want edge efects to be handled). Conv() works with any weights. Movmean only handles uniform weights as far as I know.

Answer 3

[Sulaymon Eshkabilov]

In this case, it is better to use filter() that gives a good moving average filter solution:

A = 1:5;
B = [.5 .5 ];
Out_A = filter(B, 1, A)

Out_A = 1×5

0.5000 1.5000 2.5000 3.5000 4.5000