Problem with Cluster command

When I ran the following three commands from the Statistics and Machine Learning Toolbox in the shown order, in my command window

Y = pdist(X)

Z - linkage(Y)

T = cluster(Z,'cutoff',1.2)

where X is 3042x53 matrix of type double, I get the following error (just after I execute statement 3) "Index in position 1 is invalid. Array indices must be positive integers or logical values." I then repeated these commands with a smaller X input (4x8). Same result. I believe that "cluster" wants integers for the first two columns of the Z input, (Z was produced by linkage) but the last column of Z is a distance measure and is real - so all columns of Z must be real. What am I doing wrong or misunderstanding?

Answer 1

[Walter Johnson]

Well thanks. It's something I will be looking out for now

Answer 2

[Walter Roberson]

Z = linkage(squareform(Y)) ;

[Walter Johnson]

Thanks Walter, but when I tried this I received the same error message when I ran the cluster command.

[Walter Roberson]

X = rand(4,8);
Y = pdist(X);
SY = squareform(Y);
Z = linkage(SY);
T = cluster(Z, 'cutoff', 1.2)

works for me. Tested in R2019a.

What shows up when you use

which linkage

? You should see toolbox/stats/stats/linkage.m . I am wondering if perhaps you are accidentally getting the linkage() routine from Peter Corke's Robotics Toolbox.

[Walter Johnson]

Still does not work, I copied and ran your code (see below). The linkage routine looks right too. Very strange.

>> X = rand(4,8);

Y = pdist(X);

SY = squareform(Y);

Z = linkage(SY);

T = cluster(Z, 'cutoff', 1.2)

Index in position 1 is invalid. Array indices must be positive integers or logical values.

>> which linkage

C:\Program Files\MATLAB\R2019a\toolbox\stats\stats\linkage.m

>>

[Walter Roberson]

Hmmm... I just realized that you might somehow have an existing variable named cluster

[Walter Johnson]

Ha, Bingo! Thanks a lot. Did you guess this because it happens fairly often? I feel pretty dumb.

[Walter Roberson]

It does happen a fair bit, most commonly with a variable named sum .