Assign Ranking in Matlab

I have several cell-type variables A{Y,1}(Z x 13 cells) all sorted by column C. I need to assign a ranking (add column R) for the corresponding sorting order.

Example for a B {Y,1} (8 x 3 cells). My original cell is:

A   B      C        
MJ 65   0           
MJ 321  0,0125      
MJ 2    0,0125      
MJ 1987 0,0125    
MJ 87   0,02        
MJ 5    0,0375      
MJ 743  0,0375      
MJ 124  0,05

I would like to rank (column R) each A{Y,1} cell-type, considering that in case of tie, the mean value should be assigned.

A   B      C        R
MJ 65   0           1
MJ 321  0,0125      3   %Assign mean value
MJ 2    0,0125      3
MJ 1987 0,0125  >>  3
MJ 87   0,02        5
MJ 5    0,0375      6,5 %Assign mean value
MJ 743  0,0375      6,5
MJ 124  0,05        8

Thanks for your help.

[Azzi Abdelmalek]

This is not clear

[Maria]

Ok I will edit it. Thanks

[Maria]

Do you think it's more clear now? If not, I will just try to explain it in a different way!

[Azzi Abdelmalek]

How did you get the ranking?

[Maria]

I didn't! That's what I want to get!

[Maria]

I just tried to post an example that explains what I would like to get at the end!

[Azzi Abdelmalek]

I can't understand what you want

[Maria]

I just edited it again. Please let me know if you at least understand. If not it's ok, it's my problem! Thanks :)

[Azzi Abdelmalek]

Assign mean values of what?

[Maria]

When doing the ranking, if you have in column C 3 values that are equal, you assign the mean value to each one. Meaning in this case the mean value of 2,3 and 4 is 3 (for rows 2, 3 and 4). The mean value of 6 and 7 is 6,5. This way i get a fair ranking.

Answer 1

[Azzi Abdelmalek]

   v={'MJ' 65   0           
'MJ' 321  0.0125      
'MJ' 2    0.0125      
'MJ' 1987 0.0125    
'MJ' 87   0.02        
'MJ' 5    0.0375      
'MJ' 743  0.0375      
'MJ' 124  0.05}
[a,b,c]=unique([v{:,3}])
d=accumarray(c,(1:numel(c))',[],@mean)
v(:,4)=num2cell(d(c))

[Maria]

I have a problem here! My v is a cell within a cell so how can I integrate it in your code? For instance 'v = V(:,1);' gives the error in the second line that 'Index exceeds matrix dimensions'!

[Azzi Abdelmalek]

Can you give a short example

[Maria]

For instance, if I run the code like this:

v=V{:,1};
[d,e,f]=unique([v{:,13}]);
g=accumarray(f,(1:numel(f))',[],@mean);
v(:,14)=num2cell(g(f));

It works pefectly,but it's only for the first cell of the cell!

[Azzi Abdelmalek]

attach a mat file containing your cell array

[Maria]

it adds to V{1,1} (27x13 cell) a 14th column with the ranking! But I have until V{2000,1}. And I tried different ways and still not working! I will continue trying!

[Maria]

I am not being able to attach mat file, I am attaching a pdf file. With the code and two images. Hope it's ok.

Answer 2

[Joseph Cheng]

Well we first start with finding which ones are the same.

 C = [0 .0125 .0125 .0125 .02 .0375 .0375 .05];
 R = 1:length(C)
 dC = diff(C);
 eRank = find(dC ==0);
 eRank = unique([eRank eRank+1]);

With eRank i have isolated the same valued C values. Now to see which ones are grouped together.

 eRankSpacing = [0 find(diff(eRank)>1) length(eRank)]

Now to substitute the averages of the consecutive same value ranks.

for i =1:length(eRankSpacing)-1
    tempave = mean(eRank(eRankSpacing(i)+1:eRankSpacing(i+1)));
    R(eRank(eRankSpacing(i)+1:eRankSpacing(i+1)))=tempave
end

[Joseph Cheng]

I left it without doing it inside cells just so it appears clear on what was done without the added complexity of dealing with cells.

[Maria]

Yes I noticed it. Thanks, just one question, if my C is collumn 9 of a cell within a cell. How can I translate it to the code you wrote? What I have done 'C = B{:,1(:,9)}' is naturally not working! :|

[Joseph Cheng]

I think it may be C = B{:,1}(:,9)

such that

   X = [{magic(3)},{magic(5)},{magic(10)}]
   Y = [{X}, {X},{X}];

I need to call

   Y{3}{2}(:,1)

to get 3rd cell of Y then 2nd cell within that one then all rows from column 1.

which should give me a 5x1 (in this example of magic)

[Maria]

I tried it's not, it gives the error 'Bad cell reference' (just in the 1st row). I'll continue searching it in the internet!

[Joseph Cheng]

Well maybe Azzi's solution will work.

[Maria]

I am trying it! Thanks :)

[Joseph Cheng]

how about breaking it out of the cells and put them into arrays. Then putting them back? Not optimal but doable? I like Azzi's method but sometimes the long (non-built in function) way gets you thinking how these things are performed.