missingness and PCs

missingness_pcs.R

impute <- function(X)
{
	apply(X, 2, function(x) {
		x[is.na(x)] <- mean(x, na.rn=T)
		return(x)
	})
}


# Original matrix
a <- matrix(rnorm(1700*20000), 1700)


# Missing at random
b <- a
b[sample(1:length(a), 500)] <- NA
b_imputed <- impute(b)


# Non-random missingness, e.g. take the highest value of every column
c <- apply(a, 2, function(x) {
	x[which.max(x)] <- NA
	return(x)
})
c_imputed <- c
c_imputed[is.na(c_imputed)] <- 0



# PCs
a_pc <- prcomp(a)
b_pc <- prcomp(b_imputed)
c_pc <- prcomp(c_imputed)


diag(cor(a_pc$x[,1:10], b_pc$x[,1:10]))
diag(cor(a_pc$x[,1:10], c_pc$x[,1:10]))