# How to fix a shift issue after taking hamonic mean?

1 view (last 30 days)
Nisar Ahmed on 24 Aug 2022
Commented: Mathieu NOE on 12 Sep 2022
Hi everyone,
In the figure below, the red curve (Mdry) is the harmonic average of blue curve Md by using moving mean of 4 values. Due to harmic mean the red peak has been shifted a bit as shown in figure. How can I fix it? I have attached the data of both Md and Mdry. I am using following code to create a haromic mean (in such a way that the length of both Md and Mdry should remain same -- at the end).
Mdry = zeros(size(Md));
for k=1:length(Md)-3
Mdry(k)=harmmean(Md(k:k+3));
end
k = length(Md)-2;
Mdry(k) = harmmean(Md(k:k+2));
k = length(Md)-1;
Mdry(k) = harmmean(Md(k:k+1));
k = length(Md);
Mdry(k) = Md(k);

Mathieu NOE on 2 Sep 2022
hello
is there a reason why the harmonic averaging is needed
here I compared it with a standard smoothing function with gaussain window (odd length) - yellow trace
seems to get similar results in terms of smoothing while better keeping peaks positions
Mdry = zeros(size(Md));
for k=1:length(Md)-3
Mdry(k)=my_harmmean(Md(k:k+3));
end
k = length(Md)-2;
Mdry(k) = my_harmmean(Md(k:k+2));
k = length(Md)-1;
Mdry(k) = my_harmmean(Md(k:k+1));
k = length(Md);
Mdry(k) = Md(k);
% my suggestion
Mdry2 = smoothdata(Md,'gaussian',7); % NB : odd window length to keep peak position closest to original position
plot(Md)
hold on
plot(Mdry)
plot(Mdry2)
hold off
function out = my_harmmean(in)
out = numel(in)./(sum(1./in));
end
Mathieu NOE on 12 Sep 2022
ok - no problem; take your time !
all the best

### Categories

Find more on Spectral Measurements in Help Center and File Exchange

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by