How to find the indices of same values within a vector and bin them together

I have two vectors with the same length (N). For example:

A=[ 1 2 3 3 3 3 4 5]; B=[ 9 8 7 6 5 6 7 8];

What I need to do is to take an average over elements in vector 'B' that their corresponding indices in vector 'A' have same values. In this example, element 3,4,5,6 in vector A are all equal to '3' so I need to take an average over (7+6+5+6/4) in vector B. I need to do this over a large data and bin them together. I tried different ways and it didn't work. How can I do this?

[Ameer Hamza]

What about other elements (i.e. 1, 2, 4 and 5) in A. What is your expected output for the given A and B. What have you tried so far?

[MatlabUser17]

I want to keep the other elements as well. Obviously, since the other elements have not been repeated in this example the average over one element is the same, but if they repeat I will take an average of them as well.

In summary, I need to find the same values from vector A and bin them together and take an average over their corresponding values in vector B.

let's say I have a large data:

kk=find(~isnan(B)); A=A(kk); B=B(kk); [A,IX]=sort(A); B=B(IX); u=unique(A) hist1=histc(A,u)

Up to here I have sorted the data from small to large, and have found similar values in A with their number of repetition. But I need to find their indices so I can extract them from B and take the average.

[Ameer Hamza]

Please look at my answer below. Is this what you are trying to get?

Answer 1

[Ameer Hamza]

The following statement will return the list of all the averages according to the way you mentioned in your question.

meanVector = splitapply(@(x) mean(x), B', A')

since A have 5 unique elements, meanVector will also contain 5 elements corresponding to the mean value from B.

[MatlabUser17]

Thanks but this didn't work. In Y = splitapply(func,X,G) command, my problem is defining G. G is not 'A' in my example.

[Ameer Hamza]

What are X and G? In your question, you only mentioned A and B. What is size of X and G matrix? What is not working?

[MatlabUser17]

Sorry for the confusion. X=B, and G=unique(A), but then they wont have the same number of rows!

[Ameer Hamza]

I never told you to apply unique() function on A. try exactly following the statement as I wrote in my answer.

[MatlabUser17]

This is the error I get when I apply your answer:

Error using splitapply (line 111) For N groups, every integer between 1 and N must occur at least once in the vector of group numbers.

By the way, I see the function mean(X), but where do you define mean over same values in vector A?

[Ameer Hamza]

The A vector you posted in question contained all values from 1 to 5. Therefore I was under the impression that A vector will not miss any value. But if it is the case, you need this

meanVector = splitapply(@(x) mean(x), B, findgroups(A));

This will work correctly if some number from A are missing but all elements are increasing. e.g.

A = [ 1 2 4 4 4 5 9] % all elemnts are increasing order.

but the following will give unexpected output, which might be difficult to process later.

A = [1 3 3 3 2 5 9 8]; % all elemnts are not in increasing order

If this still didn't work then please give a comprehensive example of vector A.

[MatlabUser17]

Thanks, I didn't get any errors this time (sorted the data in increasing order). But would you mind elaborating a little on using the findgroups(A) command?? For example, if A is {'b','a','a','b'}, then findgroups returns [2 1 1 2]. This just shows how many of each value are within the vector. Then, splitapply should split B based on findgroup(A).

Also is the output data in the same increasing order? I mean could I do this:

A=unique(A); B = splitapply(@(x) mean(x), B, findgroups(A)); plot (A,B) ?

[Ameer Hamza]

findgroups() finds groups in vector A and assign numbers from 1 to n where n unique elements are present in A. In other words, all the common element in A will get the same number. As an example

findgroups([1 2 5 5 5 9 0])
ans = 
  [2 3 4 4 4 5 1]

You can see that the output contain all numbers from 1 to 5.

Similarly, if it is cell array, it will assign numbers to each element. All the elements which are equal will get the same number.

[MatlabUser17]

Thanks for elaborating, really appreciate it.

[Ameer Hamza]

You are welcome.

Answer 2

[Star Strider]

See if this works with your data:

A=[ 1 2 3 3 3 3 4 5];
B=[ 9 8 7 6 5 6 7 8];

[Au,~,ix] = unique(A(:),'stable');                      % Unique Elements & Indices (Convert To Column Vector), Keep Original Order
Ar_mean = accumarray(ix, B, [], @mean);                 % Means Of Single & Repeated Values
Result = [Au, Ar_mean]                                  % Elements Of ‘A’ In Column #1, Corresponding ‘mean’ Of ‘B’ In Column #2

AB_Tbl = table(Au, Ar_mean)                             % Table (Optional)

Result =
     1     9
     2     8
     3     6
     4     7
     5     8

AB_Tbl =
    Au    Ar_mean
    __    _______

      1        9   
      2        8   
      3        6   
      4        7   
      5        8

[MatlabUser17]

Thanks, Star! it works, I got similar results from both answers. I wanted to accept both answers but seems like it's not possible. I really appreciate your help.

[Star Strider]

As always, my pleasure.