Pyrite | Level 9

## univariate and correlated distribution

Hello,

Could anyone explain me how to generate two bivariates and correlated normal distributions with 5000  sample size?

1 ACCEPTED SOLUTION

Accepted Solutions
SAS Super FREQ

## Re: univariate and correlated distribution

Can you say more? The title of your post says "univariate and correlated," but your message says "two bivariates," so I am confused. Are you trying to generate X from a multivariate normal and then generate a binary response based on the X variables? Such as simulated data from a logistic model?

You can use the RANDNORMAL function to generate correlated normal data. For example:

proc iml;

call randseed(1);

N = 5000;                 /* sample size */

Mean = {1 2};           /* mean of population */

Cov = {2.4 3, 3 8.1};/* covariance of population */

x = RandNormal( N, Mean, Cov );

2 REPLIES 2
SAS Super FREQ

## Re: univariate and correlated distribution

Can you say more? The title of your post says "univariate and correlated," but your message says "two bivariates," so I am confused. Are you trying to generate X from a multivariate normal and then generate a binary response based on the X variables? Such as simulated data from a logistic model?

You can use the RANDNORMAL function to generate correlated normal data. For example:

proc iml;

call randseed(1);

N = 5000;                 /* sample size */

Mean = {1 2};           /* mean of population */

Cov = {2.4 3, 3 8.1};/* covariance of population */

x = RandNormal( N, Mean, Cov );

SAS Super FREQ

## Re: univariate and correlated distribution

I am sorry, but I do not understand.  The RMSE of WHAT goes down?  As you know, you can generate binary variables as 0/1 or as 1/0.  What do you observe if you let p --> 1-p?  In your program, that would correspond to

z[i,j] = (p<=u);

From The DO Loop