Author: Adrián Lara-Quintanilla
This function calculates a moving average along a vector. "eFave" is the
number of elements around the element in the input vector "vectorin" used
to calculate the averaged value in "vectorout". The values at the
beginning and at the end of "vectorin" that could not be calculated with
and average of "eFave" elements are calculated as an average of the
remaining values at the beginning or at the end of "vectorin".
Note: "eFave" should be an odd number. If not, '1' is added to its value.
Example: vectorin=[1 2 5 4 8 9] and eFave=4.
eFave->5, vector out=[x1 x2 4 5.6 x5 x6], note that x3 and x4 can be
calculated as a mean of the 5 elements (including themselves) around
them. Because this can not be applied on x1,x2,x5,x6, and the goal is to
smooth the series, x1=1, x2=mean(1 and 2), x3=mean(8 and 9)
If you want to know what's fast, use this:
A moving average does not require to use a for loop at all, and the runtime can be almost independent from window size.
I really appreciate your comment :), thanks!
Looping over the elements of the vector is very slow for large vectors. Then looping over the elements of the window to be averaged is much faster:
function Y = movemean(X, N)
% X is filtered along 1st dimension
% Window size is 2*N+1
[d1, d2] = size(X);
M = X;
divV = ones(d1, 1);
for i = 1:N
i2 = i + i; % Slightly faster
Z = zeros(i, d2);
M = M + [Z; X(1:d1 - i2, :) + X(1 + i2:d1, :); Z];
divV(i + 1:d1 - i, :) = divV(i + 1:d1 - i, :) + 2;
Y = bsxfun(@rdivide, M, divV);
rewrite title and description