Question

I would like to use MATLAB group statistics functions (like grpstats) on data where each observation can be in multiple groups. For example, pizzas can have {‘pepperoni’, ‘mushroom’,’onions’} or {‘pepperoni’} or whatever and then I want group stats by topping: all of the pizzas with ‘pepperoni’, all of them with ‘mushroom’, etc.

Alternatively if you know a way to do this by hand without iterating like an idiot that would also be helpful.

Answer 1

Just put repeated measures in different rows. For example:

store = repmat(cellstr(num2str((1:3)')), 3, 1);
type  = repmat({'pepperoni', 'mushrooms', 'onions'}, 3, 1);
type  = Type(:);
score = dataset({randn(9,3), 'taste', 'looks', 'price'});

data = [dataset(store, type) score];

grpstats(data(:,2:end), 'type')

Raw data:

>> data

data = 

    store      type               taste       looks       price   
    '1'        'pepperoni'        -0.19224    -0.44463    -0.50782
    '2'        'pepperoni'        -0.27407    -0.15594    -0.32058
    '3'        'pepperoni'          1.5301     0.27607    0.012469
    '1'        'mushrooms'        -0.24902    -0.26116     -3.0292
    '2'        'mushrooms'         -1.0642     0.44342    -0.45701
    '3'        'mushrooms'          1.6035     0.39189      1.2424
    '1'        'onions'             1.2347     -1.2507     -1.0667
    '2'        'onions'           -0.22963    -0.94796     0.93373
    '3'        'onions'            -1.5062    -0.74111     0.35032

Group stats:

>> grpstats(data(:,2:end), 'type')

ans = 

                 type               GroupCount    mean_taste    mean_looks    mean_price
    pepperoni    'pepperoni'        3              0.35459      -0.10817      -0.27197  
    mushrooms    'mushrooms'        3              0.09674       0.19138      -0.74791  
    onions       'onions'           3             -0.16704      -0.97992      0.072449