cor.with {NCmisc}

Simulate a correlated variable

Description

Simulate a variable correlated at level 'r' with cector x (of the same length). Can either 'preserve' the mean and standard-deviation, leave standardizeed, or select new mean 'mn' and standard deviation 'st'.

Usage

cor.with(x, r = 0.5, preserve = FALSE, mn = NA, st = NA)

Arguments

x

existing variable, to which you want to simulate a new correlated variable

r

the 'expected' correlation you want to target (randomness will mean that the actual correlation will vary around this value)

preserve

logical, whether to preserve the same mean and standard deviation(SD) as x, for the new variable

mn

optional, set the mean for the new simulated variable [must also set st if using this]

st

optional, set the SD for the new simulated variable [must also set mn if using this]

Value

return the new variable with an expected correlation of 'r' with x

Author(s)

Nicholas Cooper

References

http://www.uvm.edu/~dhowell/StatPages/More_Stuff/CorrGen.html

See Also

sim.cor

Examples

X <- rnorm(10,100,14)
cor.with(X,r=.5) # create a variable correlated .5 with X
cor(X,cor.with(X)) # check the actual correlation
# some variability in the actual correlation, so run 1000 times:
print(mean(replicate(1000,{cor(X,cor.with(X))})))
cor.with(X,preserve=TRUE) # preserve original mean and standard deviation
X[c(4,10)] <- NA # works fine with NAs, but new var will have same missing
cor.with(X,mn=50,st=2) # specify new mean and standard deviation