Making a moving average filter without looking in to the future. So forecasting the data, how do you do it?

M = 11

% 4. Moving average filter

for n = 1:M

yma(n) = average(y([1:n+M-1]));

end

for n = M+1:length(t)-M

yma(n) = average(y([n-M:n+M]));

end

for n = length(t)-M+1:length(t)

yma(n) = average(y([n-M:length(t)]));

end

yma = yma';

[Mathieu NOE]

hello

can you clarify the question

do you need to extrapolate from the MA filtered data ?

[Kaz Van Rijsewijk]

Hello Mathieu,

This is the question that I have to answer and also apply to my script:

If we want to filter in real time we can't filter this way, we can only do it backwards.

24. Adjust the moving average filter accordingly, such that at the same value of M is filtered over as many points as before. Is there a phase delay now?

[Mathieu NOE]

sorry ; it's late here and I must be tired ,

I don't understand the question 24

Answer 1

[Steven Lord]

Use the movmean function with the M = movmean(A, [kb kf]) syntax given on its documentation page. Choose kf = 0 to include 0 elements forward of the current element.

x = randi(10, 1, 5)

x = 1×5

3 6 9 6 8

M = movmean(x, [1 0])

M = 1×5

3.0000 4.5000 7.5000 7.5000 7.0000

N = 4;
check = (x(N-1)+x(N))/2 - M(N)

check = 0

Note that the expression for check (the moving mean for element 4) does not include x(5), so the moving mean computation didn't "look into the future".

[Kaz Van Rijsewijk]

Hello Steven,

Thank you! But just be sure does this still apply to next question. My first question was a bit unclear I think.

• If we want to filter in real time we can't filter this way, we can only do it backwards.

24. Adjust the moving average filter accordingly, such that at the same value of M is filtered over as many points as before. Is there a phase delay now?