Merge Rows in Table with One Categorical Variable and One Numeric; Average the Numeric

My goal is to merge all of the categories into one, and to do so via taking the mean (this is confusing: look at the tables below, it's much easier to see what I want to do). It's already really tricky to find how to do this in MatLab, but it's even trickier because my categorical variable is an array.

I would like to take the following:

Categorical Variable Value

----------------------------------------------------

["A", "B"]                                52
["A", "B"]                                65
["C", "B"]                                35
["W", "T"]                                25

And get:

Categorical Variable Value

----------------------------------------------------

["A", "B"] 58.5

["C", "B"] 35

["W", "T"] 25

Notice how the first two rows merged, and did so on the "value" column by taking the mean.

If I do the following:

string(table.CategoricalVariable) == string(["A" "B"])

I get:

ans =

5×2 logical array

1 1

1 1

0 0

0 0

But I am, at this point, at a loss for how to do this concisely with that kind of ans. It feels like there should be an elegant way to do this?

Answer 1

[dpb]

S=categorical([["A", "B"];["A", "B"];["C", "B"];["W", "T"]]);
V=[52;65;35;25];
tT=table(S,V);
[~,~,ia]=unique(tT.S,'rows');
tT.G=ia;
rowfun(@mean,tT,'GroupingVariables',{'G'},"InputVariables",{'V'})
ans =
  3×3 table
    G    GroupCount    Var3
    _    __________    ____
    1        2         58.5
    2        1           35
    3        1           25
>>

It would be simpler if you created the categorical variable as

>> S=categorical(join([["A", "B"];["A", "B"];["C", "B"];["W", "T"]]))
S = 
  4×1 categorical array
     A B 
     A B 
     C B 
     W T 
>> 

instead if it is intended to be used as above; then it could be the grouping variable itself instead of creating another copy to use. groupsummary could be used instead with same caveat; you must create a grouping variable in some fashion; the simplest would be to make it a vector in the beginning. I've not come across a case where having more than a vector for a categorical variable was of use; I'm sure there's bound to be some place where that is the case, but it doesn't come to mind otomh what/how that would be. Mayhaps in 2D binning...