MATLAB: How to do boxplot for vectors with different sizes

boxploterrorstatisticsStatistics and Machine Learning Toolboxvector

I have four vectors (A1, A2, A3, A4) which have different length, I am trying to generate boxplot for these data:
boxplot([A1, A2, A3, A4], 'notch', 'on', 'color', [0 0 0], 'outliersize',0, 'labels',{'data1', 'data2', 'data3', 'data4'});
but I have an error: Error in boxplot (line 286) [groupIndexByPoint,groupVisibleByPoint,labelIndexByGroup,gLevelsByGroup,…
Does anyone know what is wrong?

Best Answer

You are passing in all of your data as one big long row vector (that is A1,...,A4 are all row vectors and you concatenate to a larger row vector with [A1,A2,A3,A4]), but boxplot treats different columns as different variables, or in the case of a row vector it will treat this as many observations from a single variable. This is fine (excepting that it does not have the results you want) but you then provide 4 labels and thus the error. For this reason you need to pass in a grouping variable as well, which will describe which group the data came from:
    G = [ones(size(A1))  2*ones(size(A2)) 3*ones(size(A2)) 4*ones(size(A2))];
    X = [A1, A2, A3, A4];
    boxplot(X,G,'notch','on','colors',[0 0 0],'symbol','','labels'{'data1','data2','data3','data4'});
Note, if you want to get rid of the outliers you want to set the symbol to the empty string as opposed to setting the outliersize to 0.
Related Question