Hi everyone, This might be a very easy question to a lot of the people here. I'm using a package called "Heatplus" to generate heatmaps using unsupervised clustering using function annHeatmap2 it gave me sth like this: https://photos.app.goo.gl/DLgaNW9pjJPaqzzW7 now my supervisor and I have difficulty understanding why C3 is considered closer to C2 and C4 rather than C1. We are not very advanced in math. Atom
•
written
by
jianceatom.wang •
Your graphic is cut off - so we can only see what's happening in the top portion of your heat map - but assuming the pattern is preserved in the unseen portion, the simple answer is that the averages of the columns making up C3 are more similar to those of C2 and C4 than to C1, so the dendrogram was drawn to reflect that. Examine the distances of the vertical lines to see that C1 exhibits distance between itself and the other cluster. Visually, remember that the clusters can be rotated. Thus it would be possible to draw the map in the order of C1, C3, C4, C2, which might make more sense to you visually, but the dendrogram distances would not change. Given that you're working in R, you could try examining the dendrogram, pull out some the columns, calculate distances and build up your intuition.
Thanks for answering! Could you please elborate a little bit on the "clusters can be rotated" part? How may i get more information on this?
Log in to answer this question.
Which clustering method was used? Euclidean distance is only the distance measure but there is also the so called agglomeration method, e.g. ward, complete, average, mcquitty...which is it?
Hi thank you for your reply. The settings are default but likely complete as it uses the default hclust function