non-conformable arrays error when multiplying two matrix (%)

Question

The next reproducible code generates 394 observations of 183 random normales, and tries to correlate them with Cholesky decomposition:

Generate the parameters

d <- 211

l <- 183

m <- -0.006495094

vectorsd <- rep(0.29, 183)

Generate random normals uncorrelated

rnormd <- as.data.frame(rnorm(l, mean = m, sd = vectorsd))

for (i in 1:(d+l))  {
  rnormd[,i] <- rnorm(l, mean = m, sd = vectorsd)
}

Generate a random semidefinite positive matrix of correlations

v <- runif(183,0.6,0.8)
corr <- `diag<-`(tcrossprod(v),1)

Generate cholesky matrix

cholesky <- chol(corr)

Correlate the normals and transpose the output

rnormd <- t(t(rnormd)%*%cholesky)

In this last instruction I get the error

Error in rnormd * cholesky : non-conformable arrays

At first a thought that the problem was going to be solved transposing my cholesky matrix, but then I realized that chol() function already transposes it.

Can anyone help me?

t(as.matrix(rnormd)) %*% cholesky works fine. You need to ensure (1) to use matrix multiplkication %*% not (*), and (2) that ncol of A equals nrow of B in A %*% B — dww
@dww So, the problem is that it is not coerced automatically to a matrix format. It worked. upvoted — Mauro
NP. BTW why did you ever convert to data.frame in the first place? Seems the whole calculation is better done with matrices — dww
voting to close as a typo, since was simply down to performing matrix multiplication on a non-matrix — dww

Roverof Roverof · Accepted Answer · 2020-01-31T14:48:27

If you want multiply two matrices you should use %*% instead of * sign. So here is the part you should fix.

rnormd <- t(t(rnormd)%*%cholesky)

non-conformable arrays error when multiplying two matrix (*%*)

1 Answers

non-conformable arrays error when multiplying two matrix (%)