In a post way back in 2007 (https://www.stata.com/statalist/arch.../msg00541.html), Scott Merryman showed how to generate two new variables with a prespecified correlation to each other and to an existing variable. The solution, using mata, is as follows:

Code:
clear
matrix C = (1, 0, 0\0, 1, 0\0, 0, 1)
corr2data x1 x2 x3, n(100) corr(C) clear

//Impose correlations:
//corr(x1,x2)=.25
//corr(x1,x3)=.33
//corr(x2,x3)=.75

mata
  P =(1, .25, .33\.25, 1 , .75 \ .33, .75 , 1)
  A = cholesky(P)
  X= st_data( ., .)
  R = X*A'
  corr = correlation(R,1)
  st_matrix ("corr", corr)
  st_matrix ("R", R)
end

svmat R
corr R*

I'm in a similar situation, except I need to simulate a single variable with a pre-specified correlation to two existing variables. Is it possible to modify the mata code to handle that scenario?

IYH