MATLAB Answers


Grouping multivariant data for Andrewsplot

Asked by xenon99942 on 2 Oct 2017
Latest activity Commented on by David J. Mack on 4 Oct 2017
Hi all!
i want to use the andresplot function with grouping specification:
andrewsplot(dspk,'Standardize','on', 'group' ...
Where dspk is a 2500x6 double array. In the documentation it is said that group can consists of either numerical or e.g. a char vector. I want every column of dspk as a own category. so in fact at the end it should be 6 categories, so i created a char vector with 6 names for that categories. But matlab wants "The 'group' parameter value must have one row for each row of X." I tried to transpose the array, but then the plot doesnt make any sense. So transposing is no possibility. Can you help me with this? thanks!


Sign in to comment.

1 Answer

Answer by David J. Mack on 3 Oct 2017
Edited by David J. Mack on 3 Oct 2017
 Accepted Answer

So andrewsplot states that "The rows of X correspond to observations, the columns to variables". It seems that you have instead used the columns to present the groups of observations (e.g. nObs x nGroups instead of nObs x nVars).
Assuming x is indeed nObs=2500 x nGroups=6, the following code will group your data correctly:
Since you have only one variable with 2500 observations per six groups,
Creates the required nObs=2500*6 x nVars=1 array.
Creates a nObs=2500*6 x 1 group array with the group members [1...1 2...2 3...3 4...4 5...5 6...6]' etc. (each group block repeated 2500 times). You might replace that with your grouping assignment, e.g.:
Greetings, David


Show 1 older comment
Oh i matlab version of 2011 doesn't have the function repelem in its function browser. Any alternatives?
I find a solution thanks! i tried it with kron(v,ones(1,3)) were v is the vector which one would like to repeat elementwhise
Hey xenon! My bad! I forgot to add that repelem is relatively new addition in Matlab R2015a. In older versions and as a more readable (and slightly faster) alternative to kron use repmat:
g = reshape(repmat(groupNames,size(x,1),1),[],1);
where groupNames is a row vector and x is defined as above. The reshape creates the column vector as in the example above. An additional advantage of repmat is, that you can repeat arbitrary arrays (e.g. also cellstrings with the group labels).
Greetings, David

Sign in to comment.