# Anyone know a good software to generate simulations [other than R]?

#### noetsi

##### No cake for spunky
Anyone know a good software to generate simulations [other than R]?

#### spunky

##### Can't make spagetti
you like Mplus, don't you? it has Monte Carlo simulation capabilities. so does EQS (which I think you acquired recently).

SAS does them, Stata does them... Minitab/SPSS up to a certain, convoluted extent do them...

#### hlsmith

##### Less is more. Stay pure. Stay poor.
You use SAS typically, so there you go. That dude with a SAS webpage, Rick Wicklin (sp?, something like that) released a book totally on simulating data in SAS. I used the Google book sample of it to simulate some basic data last week.

#### Jake

Like hlsmith said, you can do it in SAS, but IMHO this is one thing that's actually a much bigger pain to do in SAS compared to R (speaking as someone who has written many simulations in both).

#### noetsi

##### No cake for spunky
I have a philisophical opposition to using R [I hate to code unless it is SAS code I already know].

#### Dason

Indeed.

Simulations in R are a breeze. For instance here I'm simulating 1000 observations from the sampling distribution of the mean for 15 observations from the exponential distribution.

Code:
sampling_dist <- replicate(1000, mean(rexp(15)))
hist(sampling_dist)

#### Dason

But to be honest if you're doing simulations you're probably going to have to write some code. I'm sure somebody *could* create a stupid GUI to make doing simulations for certain things but it would have the same problem a lot of GUIs have - it's hard to make a menu for every possible option. Code is great because you aren't limited by what is in a menu.

So yeah - you should definitely code your simulations in C. That's the best choice.

#### spunky

##### Can't make spagetti
[I hate to code unless it is SAS code I already know].
noetsi, this is considered by many the 'bible' of SAS simulation for the social sciences. it was written by Xitao Fan (with whom I had a mini-fall out last year during AERA regarding the proper way to simulate multivariate, non-normal data) but the dude knows his stuff and, apparently, his hands-on approach to how to code simulations in SAS makes his book a very good reference:

http://www.amazon.ca/SAS-Monte-Carlo-Studies-Quantitative/dp/1590471415

#### trinker

##### ggplot2orBust
I think that you can actually conduct the simulation using marbles in a pouch.

#### Dason

I think that you can actually conduct the simulation using marbles in a pouch.
I think we can boil this down to make life even easier.

Essentially all simulations require random number generation of some sort. We can generate any distribution by transforming uniform random variables. We can generate uniform random variables from an infinite sequence of Bernoulli(.5) random variables. In practice we can just use a finite string of bernoullis. If we have a fair coin we can use that to generate bernoulli(.5) random variables.

Therefore if you want to do a simulation all you really need to do is flip a coin for a long time. This gets rid of all of the coding which I think would make noetsi very happy.

#### noetsi

##### No cake for spunky
I'm sure somebody *could* create a stupid GUI
Quit insulting GUI's bot. Just because they get all the cushy jobs and attention and robots don't is no reason for insulting them

#### vinux

##### Dark Knight
Use @risk. From our discussions and your interest, I don't think you like to do simulation using SAS. Because you need to do some level of programming (If you like proc iml, it may not be that difficult.)

At risk is an addins of excel and mostly gui based. Here is the LINK