Dividing a set of numbers into set of intervals and caculating the mean of each interval and comparing it with the previous one

Hi. I have a set of random numbers let's say x=[-2,-4,-7,-1,...,-3], and I need writting a code to divide x into equal intervals (let's say each interval has 50 numbers), then I want to caculate the mean of each interval and compare it with the mean of the previouse interval. Can any one help me with that please. your help is much appreciated in advanced.

Answer 1

[Matt J]

subs=discretize(x,50); %50 intervals

intervalMeans=accumarray(subs(:),x(:),[],@mean)

[Mohamed Saleh]

I got an error says 'Error using discretize Second input, edges, must have at least 2 elements'.

[Matt J]

Please copy/paste the full error message (do not transcribe). Also, what Matlab version are you running?

[Mohamed Saleh]

Thanks a lot for your reply. I got

 Error using discretize (line 61)
Second input, edges, must have at least 2 elements.

Error in new_test_project (line 4)
subs=discretize(x,50) %50 intervals

[Matt J]

what version of Matlab?

[Matt J]

Try this

subs=discretize(x,  linspace(  min(x),max(x),50)  ); %50 intervals

intervalMeans=accumarray(subs(:),x(:),[],@mean)

[Mohamed Saleh]

I am running R2016. Actually this command does not give the right number of intervals and does not devide x to equal 50 intervals.

[Matt J]

What evidence do you see of that?

[Mohamed Saleh]

For simplicity let x=[1,2,3,4,5,6];. Then after applying the command:

subs=discretize(x,  linspace(  min(x),max(x),3)  );.

I expected to have two intervals each interval has 3 elements, but I got the following results:

 subs =

     1     1     2     3     4     4

[Mohamed Saleh]

Sorry for bothering. I just want to have exactly the same elements of x without any change, just to divide x to certine number of intervals. Then I want to calculate the mean of each interval and compare the mean of the next interval with the mean of the previous interval.

[Matt J]

Hmmm. Well, I get the correct result,

    >>  x=[1,2,3,4,5,6];
    >> subs=discretize(x,  linspace(  min(x),max(x),3)  )

    subs =

         1     1     1     2     2     2

    >> intervalMeans=accumarray(subs(:),x(:),[],@mean)

    intervalMeans =

         2
         5

Not sure why you see what you see if this is indeed the code that you are running.

[Mohamed Saleh]

Ok thanks. But why the values of x have been changed in the above subs. I expected to have the same elements but devided to intervals.

[Matt J]

subs(i)=1 means x(i) belongs to interval #1 and subs(i)=2 means that x(i) belongs to interval #2.

[Mohamed Saleh]

Thank you very much for your help.

Answer 2

[Jos (10584)]

I understood your question somewhat different: "I have N numbers {x(1) x(2) ... x(N-1) x(N)} and want to split them into several groups of K values: {x(1) ... x(K)} { x(K+1) ... x(2*K)} {...} without not changing the order of the elements. For each group I would like to take the mean."

There are several ways to answer this question. Examples:

    % some data
    X = randi(10,1,10) % N = 10
    K = 3
    % Note that N is not necessarily a multiple of K

    idx = 1 + floor((0:numel(X)-1)/K) 
    M1 = accumarray(idx(:), X(:), [], @mean)

    M2 = arrayfun(@(ix) mean(X(ix:min(ix+K-1,end))), 1:K:numel(X))

[Mohamed Saleh]

Thank you very much for your answer.