Sort table based on number of occurrences

I have a table that I'd like to sort based on groups of a few of the columns. My table is made up of cells.

>> T=table({'A','B','C','B','B'}',{1,2,3,4,5}',{'Car','Van','Car','Car','Car'}')

T =

5×3 table

Var1 Var2 Var3

_____ _____ _______

{'A'} {[1]} {'Car'}

{'B'} {[2]} {'Van'}

{'C'} {[3]} {'Car'}

{'B'} {[4]} {'Car'}

{'B'} {[5]} {'Car'}

I'd like to sort by the number of occurences of Var1 and Var3, starting with the Var1- so since 'B' occurs 3 times, and 'Car' occurs twice with 'B', then the solution would be:

    Var1     Var2      Var3  
    _____    _____    _______

    {'B'}    {[4]}    {'Car'}
    {'B'}    {[5]}    {'Car'}
    {'B'}    {[2]}    {'Van'}
    {'A'}    {[1]}    {'Car'}
    {'C'}    {[3]}    {'Car'}

Is there someting like accumarray for strings? I tried groupsummary, but it will combine rows- Var2 is not always unique.

Answer 1

[Peter Perkins]

This has been unanswered for some time. Maybe this is still helpful.

The main problem here is cell arrays of (repeated) text and numeric. Don't use those unless you have to, and you probably don't.

>> T = table(categorical({'A','B','C','B','B'}'), ...
          [1,2,3,4,5]', ...
          categorical({'Car','Van','Car','Car','Car'}'));
>> cats1 = categories(T.Var1);
>> [~,order1] = sort(countcats(T.Var1),'descend');
>> T.Var1 = reordercats(T.Var1,cats1(order1));
>> cats3 = categories(T.Var3);
>> [~,order3] = sort(countcats(T.Var3),'descend');
>> T.Var3 = reordercats(T.Var3,cats3(order3));
>> sortrows(T,["Var1" "Var3"])
ans =
  5×3 table
    Var1    Var2    Var3
    ____    ____    ____
     B       4      Car 
     B       5      Car 
     B       2      Van 
     A       1      Car 
     C       3      Car 

[Marcus Glover]

Thanks! Always helpful! My data does come in that format and I cannot control it directly (SQL server is the source). Turning the strings into categorical variables does seem like it would solve a lot of my problems with this data and beyond.

[Marcus Glover]

Sorry to dig this up- the table creadted from my SQL pull has either 'cell' or 'double' as the datatypes. Is there a way to batch convert only the 'cell' types to 'categorical'?

Would there be a reason not to do this? Some of the cells do contain numbers- but those are for identification purposes (like a serial number) and will never be used as 'numbers'.

Thanks.

[Marcus Glover]

I fihured this out...

T=convertvars(T,@iscell,'categorical');