drawing dendrogram from pre calculated distance matrix

cluster-analysis, matrix, r

Solution

hclust needs an object of class dist. as.dist, rather than dist, should give you want you are looking for.

plot(hclust(as.dist(x)))

Problem

I know there's another post similar to this one but it has not helped my situation. I am trying to draw a dendrogram from a distance matrix I've calculated not using euclidean distance (using an earth-mover's distance from the emdist package). I am now trying to draw a dendrogram from this matrix: ``` dim(x) [1] 8800 8800 x <- x[1:10,1:10] x 1 2 3 4 5 6 7 1 0.00000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 2 0.67400563 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 3 0.02577228 0.6526842 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 4 0.37994900 0.7268372 0.1240314 0.0000000 0.0000000 0.0000000 0.0000000 5 0.85156584 1.0248822 0.6165767 0.9077611 0.0000000 0.0000000 0.0000000 6 0.51784015 0.5286874 0.5115762 0.6601093 1.1639417 0.0000000 0.0000000 7 0.19290720 0.5906327 0.6576926 0.4350795 0.2986499 0.4130357 0.0000000 8 1.57669127 1.3727582 1.4215065 1.9522834 1.0919793 0.9681544 1.0372481 9 3.01650143 3.3004177 3.0651622 3.2502077 4.1505108 2.9940774 3.6078234 10 0.48684093 0.6997258 0.3959822 0.3515030 0.8611233 0.5505790 0.3047047 8 9 10 1 0.000000 0.000000 0 2 0.000000 0.000000 0 3 0.000000 0.000000 0 4 0.000000 0.000000 0 5 0.000000 0.000000 0 6 0.000000 0.000000 0 7 0.000000 0.000000 0 8 0.000000 0.000000 0 9 3.753577 0.000000 0 10 1.500342 3.309016 0 ``` the problem is when I run ``` plot(hclust(x)) ``` I get this error: Error in if (is.na(n) || n > 65536L) stop("size cannot be NA nor exceed 65536") : missing value where TRUE/FALSE needed whereas if I run the dist function to calculate euclidean distances from the distance matrix that I've already calculated using a different approach, it draws the plot. ``` plot(hclust(dist(x))) ``` However, this is not realistic. I need hclust to work from the distance matrix I've already calculated using a different approach. Any ideas?

Original source