Statistical Software Packages

Anyone here use SAS? I currently do most of my work with STATA but my firm might look into SAS because we are running ahead of budget this year. I do run some optimization programs through excel/VBA and I heard that you can do optimizations faster with SAS along with regression analysis. I would welcome any useful feed back regarding these packages or another one that you use.

I use sas. Not a pro at it, but it can do anything I need to do and hell of a lot more.

  1. If you are running optimization programs through Excel, switch to any programming language and they will run much faster (even VBA by itself is much slower than C#, etc). 2) Running ahead of budget is a really dumb reason to buy a monster statistical system and learn how to use it. Writing SAS code is different than any other programming code and it takes awhile to get good at it. 3) SAS is a massive thing to just do regression analyses. 4) What’s wrong with Stata?

I personally find SAS to be a step down from Stata - for econometric work anyway, can’t speak to anything else.

  1. What other languages would you suggest. i have used C++ but that was back in freshman year of college. 2, 3 & 4) There is nothing wrong with stata but it is an older version, we have more users that have used SAS, we have financial flexibility this year and if SAS can provide other analysis/optimization capabilities in addition to running regressions i thought it would be an upgrade over what we currently have. I am sure it will be a time committment to learn but if it is that much better of a software package (as I am told) I thought it would be a useful addition. i am sure you have used both so do you think it is a waste to get and learn SAS? I know most other financial institutions use it so it has to be worth something.

I use SAS. They have an excellent Genetic Algorithm optimization module. It interacts perfectly with Oracle and it’s fast. It’s going to take you sometime to learn how to write SAS code, but it’s worth it. Start by studying the example problems supplied with the software and takes pieces of the codes that you might use. I agree with JDV, if it’s just to do regression or simple time series analysis, don’t bother.

contango1 Wrote: ------------------------------------------------------- > 1) What other languages would you suggest. i have > used C++ but that was back in freshman year of > college. > C++ is a fine language and since the operating system is written in it, it’s the Lingua Franca of Windows. I hate it, but live with it. I like Delphi and C#. > > 2, 3 & 4) There is nothing wrong with stata but > it is an older version, we have more users that > have used SAS, we have financial flexibility this > year and if SAS can provide other > analysis/optimization capabilities in addition to > running regressions i thought it would be an > upgrade over what we currently have. It surely will but it’s a big hammer. > I am sure it > will be a time committment to learn but if it is > that much better of a software package (as I am > told) I thought it would be a useful addition. i > am sure you have used both so do you think it is a > waste to get and learn SAS? I know most other > financial institutions use it so it has to be > worth something. I guy with a stats degree from Carolina can’t say much bad about SAS (in Cary, NC right next door) where they employ a bunch of my friends and professors. It’s great software but it’s big and unnecessary for simple stuff.

I suppose that if there is a long term plan to switch to SAS, then having a budget surplus is a decent reason to switch early. It does seem that SAS is an industry standard and I’ve been debating the value of learning it - I’m more familiar with SPSS and STATA.

When I worked at FactSet, I know alot of clients either used our optimization software, or fed the data we provided into MatLab. I forget how much it costs though but you could check it out here: http://www.mathworks.com/products/product_listing/index.html

Cary? *sticks finger down throat* Unfortunately, I’ve spent more than my fair share of time there; one big suburb.

I used SAS for pure stat work. Find Eviews is better for finance, if you’re real good and cheap go for R (it’s free) and that way you can write your own tests and models.

R is very nice, and its data model is very similar to Matlab’s. The one pain in the butt is getting data into the workspace. And, like Matlab, R isn’t so good about facilitating data lables (it’s there, just a little bit clunky). Must have been designed by mathematicians, rather than natural or social scientists.

FIAnalyst Wrote: ------------------------------------------------------- > Cary? > *sticks finger down throat* > > Unfortunately, I’ve spent more than my fair share > of time there; one big suburb. Yeah, well 20 years ago Cary was pretty rural and a nice place to be. The population was something like 25K. Now I think the population is > 100K and, as you say, suburban sprawl. There are still things to like about it though - highly educated population, fine schools, low crime, relatively affordable, great basketball nearby, etc…

I used SPlus in graduate Time Series class, really liked it.

Different kind of software…

R is amazingly powerful and has a lot of packages. also, there is GNU Octave if you wanted to get a feel for matlab.