Mean calculations in a for-loop

mean_240= zeros(2976,1);     %preallocate vector
mean_240(1)=mean(T(1:216),'omitnan');
index = 2;
for i = 217:size(T,1)
    mean_240(index) = mean(T((i-239):i),'omitnan');
    index = index+1;
end 

Hi all!

i have a 3192x1 vector and i want to calculate the 'backward' mean of every element of T ,considering the last 240 elements of each datapoint. i want to start at datapoint T(217) (the first 216 elements mean are calctulaeted seperately outside the lop), but as you can see, there are not enough values from 1:216 to calculate a 240element-mean and the above code won't work. so i want to modify the mean calculation until there are enough elements to proceed with the 240-mean-loop.

E.G. for the first for-loop step i = 217 the mean would be

mean_240(2) = mean(T((i-216):i),'omitnan');   

and for the next step i = 218 we can use one more sample point to calculate the mean

mean_240(3) = mean(T((i-217):i),'omitnan');   

.. and so on until we reach i = 240, then we above loop would work and we can calculate the whole 240point mean for the other sample points. how can i write that?

Answer 1

[xenon99942]

Hello!

i think i have a first solution. Don't know if it's espensive or not:

mean_240= zeros(2976,1);
mean_240(1)=mean(T(1:216),'omitnan');
index = 2;
for i = 217:size(T,1)
    if i<239
        mean_240(index)=mean(T((i-(214+index)):i));
    
    else 
    mean_240(index) = mean(T((i-238):(i+(index-2))),'omitnan');
    %index = index+1;
    end
    index = index+1;
end 

[xenon99942]

sorry the index = index+1 should be outside the if statement. But no i am getting still a solution matrix with only zeros in it. can you see the error?

[madhan ravi]

upload T

Answer 2

[Andrei Bobrov]

EDIT

Let A is your vector with length n = 3192.

i1 = 217;
ii = 240;

B = A;
n = numel(A);
lo = isnan(A);
B(lo) = 0;
p = ones(n,1);
p(lo) = 0;
k = ones(ii,1);
Bp = conv2([B,p],k);
out = Bp(i1:n,1)./Bp(i1:n,2);

[Steven Lord]

You can specify the 'Endpoints', 'discard' name-value pair arguments rather than post-processing to throw away the first 216 elements.

Though I am a bit confused about the disconnect in the original message between "i want to start at datapoint T(217)" and "there are not enough values from 1:216 to calculate a 240element-mean" since there aren't enough values from 1:217 to compute a 240-element mean either. So maybe 'discard' is not what the original poster wants.

[Andrei Bobrov]

Thank you Steven for your clarification.

[xenon99942]

no i don't want to discard the first 216 elements. I want to compute the "most-element"-mean, which is possible. thats why i want to increase the number of elements in each step till i reach datapoint index 240 from which i start to compute the backward 240-element-mean of every following element. I am using an old matlab version, so "movmean" is not available.

T is an 3192x1 vector of data type "double"

The above code doens't run anymore. Error message:

??? Error using ==> sum

Trailing string input must be 'double' or 'native'.

Error in ==> mean at 30

y = sum(x,dim)/size(x,dim);

Befor that error message ( the code is the same), i just got zeros as output. Can you see the error?