Aptech » Answers for "Simulating from a t distribution" http://www.aptech.com Fri, 08 Feb 2013 19:12:28 +0000 en-US hourly 1 http://wordpress.org/?v=3.4.1 By: BarbaraSadaba http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1797 http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1797#comments Sun, 30 Sep 2012 14:45:27 +0000 BarbaraSadaba http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1797 Sorry, my mistake, this is a function you are creating yourself. Please, disregard my last answer.

Best

Barbara

]]> http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1797/feed/ 0 By: BarbaraSadaba http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1796 http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1796#comments Sun, 30 Sep 2012 14:42:41 +0000 BarbaraSadaba http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1796 Thanks a lot. Basically my problem started because if you search in the help for this command that you showed me now “rndT”, is non existent.

It would be good to include it I believe.

Best

Barbara

]]> http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1796/feed/ 0 By: aptech http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1793 http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1793#comments Sun, 30 Sep 2012 14:36:15 +0000 aptech http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1793 Continue reading →]]> The simplest way to accomplish this is by using the student-t inverse cdf to transform random uniform numbers like this:

proc (1) = rndT(rw, cl, df);
	retp(cdftci(rndu(rw, cl), df));
endp;

You can test that the distribution of your numbers is correct by checking to see what percentage of your numbers are less than a certain value and then comparing that with the T-CDF function:

rw = 1e5;
cl = 1;
df = 5;
r = rndT(rw, cl, df);
checkVal = 0.37;

comp = cdftc( checkVal, df);

//Since cdftc returns the compliment subtract 1
print (1 - comp);
print (sumc(r .<= checkVal))/rw;

proc (1) = rndT(rw, cl, df);
    retp(cdftci(rndu(rw, cl), df));
endp;

You can copy and paste the code snippet above to check. With a ‘checkVal’ of 0.37 both print statements should be about 0.636. You can change ‘checkVal’ to test other points.

I think there may be a faster method to make the random deviates in terms of the function rndGamma. But, I don’t know it off the top of my head.

]]> http://www.aptech.com/questions/simulating-from-a-t-distribution/#answer-1793/feed/ 0