How to separate table columns by groups?

If I have a table of data (T1) such as:

Var1 Var2
1     xyz
1     xyy 
2     xxx
2     xzx
3     xyy
4     yzy
4     xzz
4     yzz

I would like to output another table (T2):

Var1      Var2
1        xyz,xyy
2        xxx,xzx
3          xyy
4      yzy,xzz,yzz

I have unsuccessfully used unstack, since it creates the values in Var2 as column names. Any suggestions would be much appreciated.

[dpb]

Is Var2 numeric?

[Denis Pesacreta]

No it is not. It is categorical of the form 'xyz'.

[Denis Pesacreta]

Var 1 is the output of a findgroups command, Var2 is categorical. Such as if Var1 was the output of a findgroups command on an ID variable, and Var2 was possessions (i.e person number one has a car and a house, while person number 4 may have a car, a house, and a computer)

Answer 1

[Guillaume]

If Var2 is a categorical array, then I assume that your xyz,xyy is actually a 1x2 categorical array [xyz, xyy], not a char array. If so:

T1 = table([1;1;2;2;3;4;4;4], categorical({'xyz';'xyy';'xxx';'xzx';'xyy';'yzy';'xzz';'yzz'}))
varfun(@(values) {values'}, T1, 'GroupingVariables', 'Var1')

[Denis Pesacreta]

Thank you! The table outputted is correct, however it says 1x11 categorical instead of displaying the values. Is there any way to fix this?

[Guillaume]

No, that's the built-in disp of a table, you can't change that. tables are not really designed to hold arrays in columns. It's more designed to have scalar values in each column.

[Peter Perkins]

Guillaume, you are correct that it's sometimes difficult to display multi-column variables in a table, but tables absolutely are designed to support them (and in fact it's the reason why "variables" in tables are called "variables", not "columns" in the doc). The display works as you would expect in cases with only a few columns in each variable.

>> table(rand(3,2),rand(3,5))
ans =
  3×2 table
           Var1                                    Var2                          
    __________________    _______________________________________________________
    0.15761    0.48538    0.42176     0.95949     0.84913     0.75774     0.65548
    0.97059    0.80028    0.91574     0.65574     0.93399     0.74313     0.17119
    0.95717    0.14189    0.79221    0.035712     0.67874     0.39223     0.70605

Actually, in your solution, the second variable in the result from varfun is in fact a cell array, each cell containing a categorical vector. There's nothing at all wrong with that, it's completely supported by tables, and in fact if the different groups have different numbers of rows in the original table it's pretty much necessary. But you are right that the display is not as informative as it could be in some cases.