# Weighted mean by groups

18 views (last 30 days)
Maximilian Schwarz on 28 Sep 2022
Edited: dpb on 28 Sep 2022
I am trying to get the weighted mean of values by group id but I don't undersnatnd why my ouput (Mean) is a matrix and not a vector.
id = { 'a'; 'a'; 'b'; 'b'; 'c'; 'c'};
values = [1;2;3;4;5;6];
weights = [0.5;0.7;0.3;0.4;0.5;0.1];
exampletable = table(id, weights, values);
G = findgroups(exampletable.id);
wmean = @(w,v)w'.*mean(v);
Mean = splitapply(wmean, exampletable.weights, exampletable.values, G)
id = { 'a'; 'a'; 'b'; 'b'; 'c'; 'c'};
values = [1;2;3;4;5;6];
weights = [0.5;0.7;0.3;0.4;0.5;0.1];
exampletable = table(id, weights, values);
G = findgroups(exampletable.id);
wmean = @(w,v)w'.*mean(v);
Mean = splitapply(wmean, exampletable.weights, exampletable.values, G)

dpb on 28 Sep 2022
Edited: dpb on 28 Sep 2022
It's the wmean function definition that's doing it -- what you intended was
wmean = @(w,v)mean(w.*v);
trying that,
id = { 'a'; 'a'; 'b'; 'b'; 'c'; 'c'};
values = [1;2;3;4;5;6];
weights = [0.5;0.7;0.3;0.4;0.5;0.1];
exampletable = table(id, weights, values);
G = findgroups(exampletable.id);
Mean = splitapply(wmean, exampletable.weights, exampletable.values, G)
Mean = 3×1
0.9500 1.2500 1.5500
OBTW, you can do things like this without the explicit step of findgroups with groupsummary (or several other ways, as well)--
gmeans=groupsummary(exampletable,'id',wmean,{"weights","values"})
gmeans = 3×3 table
id GroupCount fun1_weights_values _____ __________ ___________________ {'a'} 2 0.95 {'b'} 2 1.25 {'c'} 2 1.55
Maximilian Schwarz on 28 Sep 2022
Thanks!