Anovan random model with unbalanced data set

Hello, I'm running a 2-way random model anova with unbalanced data using anovan. I know anovan can deal with unbalanced data, but I'm confused as to how. When I run my code, the empty cells are replaced with '0', but those zeros are used in calculating that group mean. I therefore end up with an incorrect value for that group and wrong statistics. I have tried putting NaN in those empty cells but then none of the data for that subject is used. Thoughts? Thanks, Jen

data_ANOVA = data(:,col_1:col_2);
data_ANOVA = data_ANOVA(:);

% Factor A, Speed
  numA  = (col_2 - col_1)+1;
  speed = repmat(1:numA, r,1);
  speed = speed(:);

   % Factor B, Subject (random factor)
    numB = 10;     % number of reps per subjects

    subject = [ones(numB,numA); 2*ones(numB,numA); 3*ones(numB,numA); 4*ones(numB,numA); 5*ones(numB,numA); 6*ones(numB,numA); 7*ones(numB,numA); 8*ones(numB,numA); 9*ones(numB,numA); 10*ones(numB,numA); 11*ones(numB,numA); 12*ones(numB,numA); 13*ones(numB,numA); 14*ones(numB,numA); 15*ones(numB,numA); 16*ones(numB,numA); 17*ones(numB,numA)];
    subject = subject(:);

    [p, table, stats] = anovan(data_ANOVA,{speed subject},'model',2,'random',2,'varnames',{'Speed';'Subject'});