topic Re: Monte Carlo and Bootstrap Simulation in Statistical Procedures

Monte Carlo and Bootstrap Simulation

mickir1 — Thu, 14 Dec 2017 17:59:50 GMT

I don't know the code for gamma distribution. I started off with this below, but I don't know where to go from there.

data proc iml;
sims=1000; *this is for the outer loop that says how many times the simulation is repeated;
size= 20; *this is the size of each simulated sample;
a=3;*parameters needed;
do i=1 to sims;
free sample;
do k=1 to size;
sample=RAND('GAMMA',a);
end;
print sample;
end

Re: Monte Carlo and Bootstrap Simulation

PaigeMiller — Thu, 14 Dec 2017 18:05:44 GMT

We don't do your homework for you.

If you go ahead and generate code to attempt to solve this problem, and it doesn't work and you can't figure out why, then we can help you fix your code and even suggest better approaches.

Re: Monte Carlo and Bootstrap Simulation

mickir1 — Thu, 14 Dec 2017 18:10:06 GMT

I’m not asking you to do my homework, I’m asking for help figuring out the gamma distribution part.

Re: Monte Carlo and Bootstrap Simulation

PaigeMiller — Thu, 14 Dec 2017 18:28:18 GMT

@mickir1 wrote:
I’m not asking you to do my homework, I’m asking for help figuring out the gamma distribution part.

That's an extremely broad question.

Can you be much more specific about what you don't understand?

Re: Monte Carlo and Bootstrap Simulation

ballardw — Thu, 14 Dec 2017 18:31:23 GMT

@mickir1 wrote:
I’m not asking you to do my homework, I’m asking for help figuring out the gamma distribution part.

Are you getting errors? Post the code and errors from the log into a code box opened with the forum {I} menu icon. The code box is important to retain formatting of error messages that may provide hints.

Re: Monte Carlo and Bootstrap Simulation

ChrisHemedinger — Thu, 14 Dec 2017 18:42:08 GMT

If you haven't already, you should review Rick Wicklin's articles about simulation, including this one:

https://blogs.sas.com/content/iml/2014/04/30/simulating-from-the-inverse-gamma-distribution-in-sas.html (that's inverse Gamma, I know)

And the whole collection of them here:

https://blogs.sas.com/content/tag/simulation/

Also...your code, as shared, is incorrect (syntax errors). This creates the sample you want, I think.

data sample (keep=x);
	sims=1000;
	size= 20;
	a=3;
	do i=1 to sims;
		do k=1 to size;
			x=RAND('GAMMA',a);
			output;
		end;
	end;
run;

proc sgplot data=sample;
 histogram x;
run;

Re: Monte Carlo and Bootstrap Simulation

PeterClemmensen — Fri, 15 Dec 2017 10:23:28 GMT

3 things:

1) You write "data proc IML" in the top of your code. But from your syntax and your note "PROC IML" in your assignment, I'm guessing you want to use PROC IML. In that case, whenever you can vectorize instead of using a loop, vectorize instead of using a loop. Like this

proc iml;
   sims=1000;
   size=20;
   SimM=j(size, sims, 0);
   call randgen(SimM, "Gamma", 3);
quit;

This gives you a 20x1000 matrix of random variates from the Gamma distribution. This gives you a nice basis to calculate the statistics you want.

It is not due until monday, so you've got plenty of time to......

3) Consult @Rick_SAS' blog The DO Loop. Also, his book Simulating Data With SAS is filled with theory and examples on the topic.

Good luck 🙂

Re: Monte Carlo and Bootstrap Simulation

Rick_SAS — Fri, 15 Dec 2017 10:47:11 GMT

@ChrisHemedinger has provided DATA step code that will generate 1000 samples from the Gamma(2) distribution, but please modify his code to use

(KEEP=i x)

Each sample is size 20. The variable 'i' in the output data can be used to index each sample. For example

data sample (keep=i x);
	sims=1000;
	size= 20;
	a=3;
	do i=1 to sims;
		do k=1 to size;
			x=RAND('GAMMA',a);
			output;
		end;
	end;
run;

proc sgplot data=sample;
 WHERE i=1;   /* plot first sample */
 histogram x;
run;
proc sgplot data=sample;
 WHERE i=2;   /* plot second sample */
 histogram x;
run;

I assume your teacher taught about BY-group processing for efficiently analyzing each sample. Either that or he/she taught you to use a macro loop. Both methods are described in "Simulation in SAS: The slow way or the BY way ."_ Use whichever way was taught.