How to use accumaray to aggregate text/string based?

Hi!

I know that we can use accumaray for aggregating some value based on key variable, but can we use it to aggregate string as well? My code to aggregate number value:

a=[11;22;22;11;22;11];
b=[10;20;30;10;20;60];
c=['a';'c';'b';'a';'b';'d'];

[i, ~,n]=unique(a,'first'); %find unique
Mode = accumarray(n , b , size(i) , @(x) mode(x)); %find mode based on keys a

This will resulting Mode=[10;20]

For example, I want to use "Mode" for text based variable, i.e. resulting Text=[a,b]

Is it possible to do it? Any help will be appreciated.

Regards!

Answer 1

[Razvan Carbunescu]

accumarray is for numerics only but if c is going to only be single characters then can probably still use accumarray as chars are treated as numerics for most functions.

>> a=[11;22;22;11;22;11]';
>> b=[10;20;30;10;20;60]';
>> c=['a';'c';'b';'a';'b';'d']';
>> [i, ~,n]=unique(a,'first'); %find unique
>> res = accumarray(n , c ,size(i), @mode)

res =

    2×1 char array

      'a'
      'b'

If you want to do grouping for more general text can look at groupsummary or findgroups / splitapply workflow to use with char vectors or strings.

>> t = table(a,b,c);
>> groupsummary(t,'a','mode',{'b' 'c'})

ans =

    2×4 table

      a     GroupCount    mode_b    mode_c
      __    __________    ______    ______

      11        3           10        a   
      22        3           20        b

[Walter Roberson]

You could also look at categorical arrays.