In addition to the five listed in this title, there are quite a few other options, so how do you choose which statistical software to use?

The default is to use whatever software they used in your statistics class–at least you know the basics.

And this might turn out pretty well, but chances are it will fail you at some point. Many times the stat package used in a class is chosen for its shallow learning curve, not its ability to handle advanced analyses that are encountered in research.

I think I’ve used at least a dozen different statistics packages since my first stats class. And here are my observations:

1. The first one you learn is the hardest to learn. There are many similarities in the logic and wording they use, even if the interface is different. So once you’re learned one, it will be easier to learn the next one.

2. You **will** have to learn another one. Just accept it. If you have the self discipline to do it, I suggest learning two at the beginning. This will come in handy for a number of reasons

– My favorite stat package for a while was BMDP. Until the company was bought up by SPSS. I’m not sure if they stopped producing or updating it, but my university cancelled their site license.

– Many schools offer only a site license for only one package, and it may not be the one you’re used to. When I was at Cornell, they offered site licenses for 5 packages. But when a new stats professor decided to use JMP instead of Minitab, guess what happened to the Minitab site license? Unless you’re sure you’ll never leave your current university, you may have to start over.

– In case you decide to outwit the powers-that-be in IT who control the site licenses and buy your own (or use R, which is free), no software package does every type of analysis. There is huge overlap, to be sure, and the major ones are much more comprehensive than they were even 5 years ago. Even so, the gaps are in the most complicated analyses–some mixed models, gee, complex sampling, etc. And when you’re trying to learn a new, highly complicated statistical method is not the time to learn a new, highly complicated stats package.

For these reasons, I recommend that everyone who plans to do research for the forseeable future learn two packages.

I know, it’s hard enough to find the time to start over and learn one. Much less the self discipline. But if you can, it will save you grief later on. There are many great books, online tutorials, and workshops for learning all the major stats packages.

But I also recommend you choose one as your primary package and learn it really, really well. The defaults and assumptions and wording are not the same across packages. Knowing how yours handles dummy coding or missing data is imperative to doing correct statistics.

Which one? Mainly it depends on the field you’re in. Social scientists should generally learn SPSS as their main package, mainly because that is what their colleagues are using. You can then choose something else as a backup–either SAS, R, or Stata, based on availability and which makes most sense to you logically.

{ 36 comments… read them below or add one }

This is Prasad, I am purshing Msc (statistics with computer applications) we can give the suggestion which software is better pls me suggests

Hai folks

Basically Iam a Statistics background student completed my graduation in Statistics And now Iam pursuing my post graduation in statistics too

Iam going learn SAS is it okay for me ??

or any thing else you people are going to suggest please suggest me

Hi friends!

I’m in an agricultural institution,and most of my colleagues are using SAS & SPSS in their researches. For a change, I’m planning to use another statistical tool for my research, the JMP. Could there be any difference?

Thanks

Cisima

I have done MA economics and looking for supporting statical package for carrer betterment. Please suggest some statical package which can support to my economics degree..

Dear Ganesh,

Given that you’re from Economics background, it is suggested to go ahead with STATA or SPSS. Stata has witnessed a rapid rise in last few years ,especially all international org in India like UN office,World Bank regional office,Institute of economic growth, NCAER they all have swaped from spss to stata.

I would recommend using R not only because it is free but because it has become an industry standard. Below is a link to a very good course that will teach you how to use R. Good luck! -Lisa

https://www.coursera.org/learn/r-programming

Hey Ganesh and Lisa, thanks for posting. We have a great collection of resources for R, too, including 2 free webinars. Check them out here.

I personally did my first degree in Quantitative Economics of Makerere University, and I have always felt at home using both STATA & SPSS tools. But more preferably STATA. If you are in the field of Economic Research, Agricultural Reseach, Public health(Epidemiology), i would advise you use STATA instead of SPSS.

In contrast, I think SPSS has better procedures when it comes to using graphs.

In general, I prefer STATA to SPSS

Could anyone suggest me any site that has some good projects ( I am looking for beginners to intermediate level) that uses Stata as a tool?

Thanks

hi Paul,

Here is a link to a book that I purchased that I thought was very helpful. I do not use Stata, however, I learned a lot by reading this book. (The workflow concepts carry over to anything.) Hope this helps!

http://www.indiana.edu/~jslsoc/web_workflow/wf_home.htm

https://www.amazon.com/Workflow-Data-Analysis-Using-Stata-ebook/dp/B01GQJSGGI/ref=sr_1_1?ie=UTF8&qid=1470399079&sr=8-1&keywords=analysis+of+workflow+stata#nav-subnav

Hey Paul and Lisa,

You can also check out our list of Stata resources, which includes 2 free webinars.

I definitely prefer NCSS, though not mentioned in the article. Now the newest version pre-released even in cloud.

https://www.apponfly.com/en/application/ncss10

Hi Brandon. I am using AppOnFly as well, but I have chosen SPSS to do my analysis. They also provide a free trial of SPSS which is available for 30days: http://www.apponfly.com/en/ibm-spss-statistics-standard?EZE

For a comparison of SPSS, SAS, R, Stata and Matlab for each type of statistical analysis, see

http://www.stanfordphd.com/Statistical_Software.html

I like to use Java since it has good graphics. Therefore, my choice is SCaVis (http://jwork.org/scavis). It integrates Java and Python with superb graphics.

I am used to spss and stata for my data analysis, however today I tried adding “analyse-it” to my excel package. It really worked for me. Can I really go ahead with it?

Hi Kule,

I don’t know much about the excel plug ins (or whatever the correct software term is). As a general rule, I avoid excel for data analysis, but this add-on may be just fine.

You could also try the XLSTAT software statistical add-on. It holds more than 200 statistical features including multivariate data analysis, modeling, machine learning, stat tests and field-oriented features https://www.xlstat.com/en/

Hi Karen, nice suggestions backed with arguments!

On a different note, I wish to hear your opinion on free software… Have you, for example, had an experience with EasyReg? It seems to have much of the econometrics methods covered — by far more than I would ever imagine to use –, it’s easy to operate and is supported with PDF-files about relevant theory. What do you think? (I have currently no access to commercial software, unfortunately.)

Hi Ragnar,

Thanks! I haven’t used that software before, but I can tell you there are many good stat software packages out there. If you like using it and you’re confident that it’s accurate, go with it.

Karen

I often use r ! and sometimes work with SPSS and Excel,but at all, i prefer

to use R because i love programing and R is a wonderfull language.also R isn’t

limited! my goal idea is to create packages that cover shortage of other

softwares,and linking softwares toghether.Indeed,i like to ferret in softwares.

so,my first software is R but i hasn’t think about primiary software yet…!

so,i research about statistical softwares and decide to use STATA inside

R!

Hi Morteza, I agree: R is awesome if you love programming. But do check out Stata too. 🙂

Hi karen do you really think that R is more efficient then Stata. I think that you are right because in programming most of my fellows using R rather then Stata. So Agreed with you…….. 😀

I use R and Stata regularly. Dollar for dollar, I personally think that Stata is the most comprehensive stats package you can buy. Excellent documentation and a great user community. R is excellent as well, but suffers from absolutely terrible online documentation, which (for me) requires third party sources (read: books).

If somebody is buying you a license, then you don’t care what it costs. If someone like me has to buy a license, then to me, Stata is a no-brainer, given all the stats you can do with it.

My college eliminated both SAS and SPSS for that reason and use R for most classes. Rumor has it SAS is offering a new “college” licensing fee, but I’m not privy to that information.

Small sidebar: SAS started on the mainframe and it annoys me that it still “looks” that way. JMP is probably better (and again, expensive) but doesn’t have anywhere the capabilities if base SAS, the last time I looked.

Just my opinions.

Hi Joe,

I actually agree with you about Stata. If I were to start over, that’s what I would use, especially, as you’ve said, if you’re buying your own license.

And Stata has the *best* manuals, IMHO.

Karen

Depends on which social scientists you are talking about. I doubt you will find many economists, for example, who do most (if any) of their analyses in SPSS. If you absolutely must have a gui JMP is clearly the superior platform, since its scripting language can interface with R, and you can do whatever you please. Try searching for quantile regression in the SPSS documentation, it says the math is too hard, and SPSS cannot compute.

Hi Jeremy,

Agreed, most economists I’ve talked to use either Stata or Eviews.

SPSS also interfaces with R.

Sure, there are examples of specific analyses that can’t be done in any software. That’s one reason why it’s good to be able to use at least two.

Karen

I am SPSS and R lover…in my university they use JMP software…how should I convince them that SPSS is better than JMP…or First of all can I convince them???

Cheers

Christos

Well, I’m sure they’ll cite budget issues. But there are some statistical options in SPSS that are not available in JMP. I don’t know of any where the reverse is true, although that may just be my lack of knowledge of JMP. For example, to the best of my knowledge, JMP doesn’t have a Linear Mixed Model procedure.

Karen

When you add random effects to a linear model in JMP the default is REML. In fact the manual goes so far as to say REML for repeated measures data is the modern default, and JMP provides EMS solutions for univariate RM ANOVA only for historical reasons. JMP doesn’t do multilevel models (more than 1 level of random effects), and I don’t believe it does generalized linear mixed effects models (count or binary outcomes). I usually use Stata and R, but I keep an eye on JMP because it is a fun program sometimes. I have used it for repeated measures data by mixed model when a colleague wanted help doing it himself, where the posthoc tests where flexible and accessible, compared to his version of Stata or in R.

Thanks, Dave. That’s great to know. The last time I used JMP (which was a few years ago), REML wasn’t an option.

Yes, I agree. JMP is very straightforward and for 95% of analyses that most researchers use, entirely sufficient.

Karen

Good advice, all around. But… if you choose SPSS as your primary package, SAS has little to offer you, and vice versa. The overlap is just too great to make either a good complement to the other.

A factor to consider in choosing between the Big Two is your preferred user interface. If you don’t want to program (much) and you adore point-and-shoot interfaces, go with SPSS. If you don’t mind programming explicitly, and despise point-and-shoot interfaces SAS will make you happier.

Another factor in choosing among the Big Two is your use of structural equation models (SEMs). If you don’t use them it’s a non-issue. If you use them extensively, you should choose between EQS-like syntax (in SAS PROC CALIS) and SPSS’s AMOS. SEMs are confusing enough without worrying about converting from your preferred expression of the models into the expression your software wants.

Much better choices as a complement to one of the Big Two are Stata and some dialect of S (R, S, S-plus). Stata users say it has some very slick programming facilities. (I’m not among them, so I can’t say from experience.) The S dialects are killers for simulation studies. I benchmarked R against SAS/IML (in version 9.1) and found R was an order of magnitude faster. R is built entirely around an object-oriented programming interface. Language extensions are a snap. In my opinion bootstrap estimation is easier in R than in other languages. High resolution graphics are native to R, and (despite a lot of improvement from versions 6 to 7 to 9.1 and 9.2) not native to SAS.

I think SAS becomes an asset over SPSS when the focus is on data preparation: Merging multiple tables, accessing SQL databases, using API functions, creating canned reports, etc..

hi friends,

I am new to R.I would like to know R-PLUS.Does any know where can I get the free training for R-PLUS.

Regards,

Peng.

I believe the above comment is spam. I am not aware of the existence of R-plus; googling revealed a word for word comment on another site: http://www.talkstats.com/showthread.php/10761-free-training-for-R-PLUS.

Apologies to the commenter if this is a genuine enquiry.

Thanks, Dave.

I suspect it was real, only because there was no link back to another site (you wouldn’t believe the strange links I get). I figured it was a language difficulty, and they meant S-Plus, on which R was based.

Karen

{ 2 trackbacks }