What is the best way to count the occurrences of two table columns together?

I have a table that looks like this:

>> A = {'A'; 'B'; 'C'; 'A'; 'C'; 'C'; 'B'};
>> B = {'x'; 'y'; 'z'; 'xx'; 'z'; 'z'; 'y'};
>> T = table(A,B)

T = 

     A      B  
    ___    ____

    'A'    'x' 
    'B'    'y' 
    'C'    'z' 
    'A'    'xx'
    'C'    'z' 
    'C'    'z' 
    'B'    'y' 
    

What is the best way of getting a result like the one blelow? I essentially want to count the occurrances of column A and B., and have as output a table.

ans = 

     A       B      occ
    ____    ____    ____

    'A'     'x'     1   
    'A'     'xx'    1   
    'B'     'y'     2   
    'C'     'z'     3  
    
   

The immediate solution that comes to mind is to combine the table columns together, and then use histc, but is there a better approach that will leave me with a Table where A and B aren't combined? Or is usng a strsplit on T2.AB the way to go?

>> c = strcat(T.A, ',', T.B)

c =

  7×1 cell array

    'A,x'
    'B,y'
    'C,z'
    'A,xx'
    'C,z'
    'C,z'
    'B,y'

>> [AB, ~, J] = unique(c);
>> occ = histc(J, 1:numel(AB));
>> T2 = table(AB, occ)
T2 = 

      AB      occ
    ______    ___

    'A,x'     1  
    'A,xx'    1  
    'B,y'     2  
    'C,z'     3  

Answer 1

[Star Strider]

Try this:

A = {'A'; 'B'; 'C'; 'A'; 'C'; 'C'; 'B'};
B = {'x'; 'y'; 'z'; 'xx'; 'z'; 'z'; 'y'};
T = table(A,B);

[Tu,~,ix] = unique(T, 'rows', 'stable');
Tally = accumarray(ix, 1);
TallyTable = [Tu, table(Tally)]

producing:

TallyTable =

  4×3 table

      A        B       Tally
    _____    ______    _____

    {'A'}    {'x' }      1  
    {'B'}    {'y' }      2  
    {'C'}    {'z' }      3  
    {'A'}    {'xx'}      1  

The accumarray function is quite versatile, and applicable here.

Answer 2

[Steve Eddins]

In my previous answer, I forgot that grpstats is in the Statistics and Machine Learning Toolbox. Here is an answer using groupsummary, which is in MATLAB.

>> groupsummary(T,["A" "B"])

ans =

  4×3 table

     A      B      GroupCount
    ___    ____    __________

    "A"    "x"         1     
    "A"    "xx"        1     
    "B"    "y"         2     
    "C"    "z"         3     

[Monika Jaskolka]

Very interesting function! Unfortunately for my purposes I need something that will work before R2018a, but this is good to know.

Answer 3

[Steve Eddins]

I think grpstats will do what you want:

>> grpstats(T,["A" "B"])

ans =

  4×3 table

              A        B       GroupCount
            _____    ______    __________

    A_x     {'A'}    {'x' }        1     
    A_xx    {'A'}    {'xx'}        1     
    B_y     {'B'}    {'y' }        2     
    C_z     {'C'}    {'z' }        3     

Also, tabular text analysis will generally be more efficient and easier to perform if you work with strings instead of cell arrays of char vectors:

>> A = ["A" ; "B" ; "C" ; "A" ; "C" ; "C" ; "B" ]

A = 

  7×1 string array

    "A"
    "B"
    "C"
    "A"
    "C"
    "C"
    "B"

>> B = ["x" ; "y" ; "z" ; "xx" ; "z" ; "z" ; "y" ]

B = 

  7×1 string array

    "x"
    "y"
    "z"
    "xx"
    "z"
    "z"
    "y"

>> T = table(A,B)

T =

  7×2 table

     A      B  
    ___    ____

    "A"    "x" 
    "B"    "y" 
    "C"    "z" 
    "A"    "xx"
    "C"    "z" 
    "C"    "z" 
    "B"    "y" 

>> grpstats(T,["A" "B"])

ans =

  4×3 table

             A      B      GroupCount
            ___    ____    __________

    A_x     "A"    "x"         1     
    A_xx    "A"    "xx"        1     
    B_y     "B"    "y"         2     
    C_z     "C"    "z"         3     

[Steve Eddins]

Note that grpstats is in the Statistics and Machine Learning Toolbox. I posted a separate answer using groupsummary, which is a newer function that is in MATLAB.