在Matlab中分组和汇总

Question

我有一个简单的矩阵，在某些列中有重复值。我需要按名称和星期分组数据，并总结每周给定的价格。这里是例子：

name day week price 
John 12 12 200 
John 14 12 70 
John 25 13 150 
John 1 14 10 
Ann 13 12 100 
Ann 15 12 100 
Ann 20 13 50 

所需的输出将是：

name week sum 
    John 12 270 
    John 13 150 
    John 14 10 
    Ann 12 200 
    Ann 13 50 

有没有一个很好的办法做到这一点？我用的循环，但不知道它是做的最好的方式：

names= unique(data(:,1)); % getting unique names from data 
n=size(names, 1);   % number of unique names 
m=size(data(:,1),1);  % number of total rows 
sum=[];     % empty matrix for writing the results 
for i = 1:n    
     temp=[];   % creating temporar matrix 
     k=1; 
    for j=1:m 
     if name(i)==data(j,1)  % going through all the rows and getting the rows of 
      temp(k,:)=data(j,:); % the same name and putting in temporar matrix 
      k=k+1; 
     end 
    end 
    count=0; 
    s=1; 
    for l = 1:size(temp,1)-1  % going through temporar matrix of one name(e.g.John) 
     if temp(l,3)==temp(l+1,3) % checking if the day of current row is equal to the 
     count=count+temp(l,4); % date of the next row (the data is sorted by name 
     else      % and date) and then summing the prices 4th column 
      sum(s, 1:3)=[names(i) temp(l,3) count]; 
      count=0;    % if the days are not equal, then writing the answer 
      s=s+1;    % to the output matrix sum 
     end   
    end 
end 

Answer 1

使用accumarray。它会分组和汇总这样的值。您可以使用第三otuput参数从unique(data(:,1))得到的数字指标传递给的accumarray的subs说法。详情请参阅doc accumarray。

Answer 2

也许最简单的方法是使用GRPSTATS功能从统计工具箱。你必须在name和week以产生第一组结合：

[name_week priceSum] = grpstats(price, strcat(name(:), '@', week(:)), {'gname','sum'});