Hello Professor Alghamdi,

The R statistical program is absolutely phenomenal and far surpasses the capabilities of SPSS.

While it may be true in your experience that SPSS is the most common statistical program used in medical research, I have noticed its use dwindling within medical research in the United States (as well as the rigor of said research). The major reason for SPSS's decrease is the limitations of its syntactical programming (though the acronym stands for statistical package for the social sciences which suggests a more psychological audience).

As a result, the SAS program is now (for the most part) the preferred tool for conducting statistical modelling. To compare SPSS syntax capabilities with those of SAS is akin to having the Qur'an paraphrased to someone by a Trmpu-supporting redneck (SPSS programming) versus reading the text with the necessary fluid comprehension of Arabic in order to fully understand its teachings (SAS programming).

This may seem like superfluous rambling but bear with me here: when using SPSS syntax, a person will never truly get a full and necessary grasp of the many intricacies and meanings of statistical syntax. On the other hand, with SAS syntax, a person is able to manipulate syntactical commands with an infinite amount of options... in other words it is true statistical programming and provides a true understanding of said programming.

The reason I mention SAS is because it is very similar to R; with R you get a true programming feel and the options are endless. The fluidity of the options will allow a user to build and hone their programming skillset. Compared to SPSS, R is VASTLY superior; compared to SAS, R is only second due to publishing prowess (not as many articles use R but many use SAS). I would always recommend using R over SPSS and there is no exception to this (though I understand people like using SPSS and I'm not going to lecture them and look like some know-it-all asshole).  

One caveat however, is since R is an open-source program, there is the potential for add-on packages to have programming errors. With this being said, I have yet to find even one truly problematic issue in any of the statistical experimentation and modeling that I have been doing for the past 10 years.

Sincerely,

Logan Netzer

I've been using R for quite some time now and have never come across any concerns regarding its use. As you have already stated, R is extremely flexible and its functionality is growing every day. Also, its user base seems to be growing and I realize that statistical anlyses in high impact journals articles are often based on R.

Dear Professor Alghamdi:  I have also used R and found it to be just as useful as SPSS.  I think it should be statistically just as robust as SPSS.  I agree with Professor Volken.

Hello Professor Alghamdi,

The R statistical program is absolutely phenomenal and far surpasses the capabilities of SPSS.

While it may be true in your experience that SPSS is the most common statistical program used in medical research, I have noticed its use dwindling within medical research in the United States (as well as the rigor of said research). The major reason for SPSS's decrease is the limitations of its syntactical programming (though the acronym stands for statistical package for the social sciences which suggests a more psychological audience).

As a result, the SAS program is now (for the most part) the preferred tool for conducting statistical modelling. To compare SPSS syntax capabilities with those of SAS is akin to having the Qur'an paraphrased to someone by a Trmpu-supporting redneck (SPSS programming) versus reading the text with the necessary fluid comprehension of Arabic in order to fully understand its teachings (SAS programming).

This may seem like superfluous rambling but bear with me here: when using SPSS syntax, a person will never truly get a full and necessary grasp of the many intricacies and meanings of statistical syntax. On the other hand, with SAS syntax, a person is able to manipulate syntactical commands with an infinite amount of options... in other words it is true statistical programming and provides a true understanding of said programming.

The reason I mention SAS is because it is very similar to R; with R you get a true programming feel and the options are endless. The fluidity of the options will allow a user to build and hone their programming skillset. Compared to SPSS, R is VASTLY superior; compared to SAS, R is only second due to publishing prowess (not as many articles use R but many use SAS). I would always recommend using R over SPSS and there is no exception to this (though I understand people like using SPSS and I'm not going to lecture them and look like some know-it-all asshole).  

One caveat however, is since R is an open-source program, there is the potential for add-on packages to have programming errors. With this being said, I have yet to find even one truly problematic issue in any of the statistical experimentation and modeling that I have been doing for the past 10 years.

Sincerely,

Logan Netzer

Dr. Alghamdi, I believe Dr. Netzer's answer covers it well. I'd also like to reflect Dr. Volken's point that the "street cred" of R is well-established: Many high-impact research shops preferentially use R (or, rather, have dedicated statisticians who do), as do many large organizations with diverse data needs (http://www.listendata.com/2016/12/companies-using-r.html).

The only piece I would venture to add to this conversation is consideration of R being open source.

Open source software is not perfect. Dr. Netzer's point about conflicting functions/terminology between packages is correct. This does mean that at times one function in a given package will be masked by another in a more-recently called package (R lets you know when this happens).

The sprawling nature of R also means that it's impossible to have single set of error codes that apply to all situations. If you use R, you will eventually get cryptic error messages. Online searches or the very active online communities can very often help solve these problems, but they will likely arise more often than they do for other statistical software.

More profoundly, the "open-source-ness" of R means that a researcher does have to learn the particulars of a given package and even of how to write code. Indeed, one general criticism of open source software in general is that it is less user-friendly than closed-source software. Whether over-blown or not, this criticism pertains here, I would argue, in the steeper learning curve to learn to use R (or SAS) than to learn SPSS. It's my opinion that one reason for the popularity of SPSS is its ease of use: Many researchers can quickly learn to perform a few, well-worn analyses in SPSS, and then continue to use them for nearly all similar experiments.

Although one _can_ settle on a set of such "canned" analyses in R or SAS, neither lends itself to it, especially R. But there are those who would argue that this is a blessing in disguise for researchers who wish to conduct more versatile and current analyses. Being open source, R lets one "look under the hood" much more easily to see exactly how analyses are being conducted, ensuring fidelity to one's particular needs (or rather, the needs of one's idiosyncratic data).

Being open sources affects R in another important way. It allows statisticians to disseminate advances in analyses much more quickly: R tends to be more cutting edge. And also thus rougher around the edges.

In all, then, I would say the "open-source-ness" of R means it is more flexible, adaptable, transparent, and certainly more extensible. But it also means it requires one to pay much closer attention to what analyses one is conducting. Although publicly most would quickly agree that that last point about paying attention to analysis is important, I have to say I have known researchers who are much more comfortable clicking a few buttons to do a few tried-and-trued analyses than they are trying to learn a whole new area (in addition to their current area of expertise).

So, in my opinion, if you're the kind of researcher who isn't afraid of learning something new (and maybe a bit scary) to become better at what you're already likely doing quite well, then--yes!--R is worth the effort.

If this _still_ isn't enough, here's another. more general conversation:

https://www.researchgate.net/post/Which_is_better_R_or_SPSS

One final addition to my rambling: Although most die-hard R users access it through a simple terminal (e.g., bash or fish), I suggest that the transition to R is a bit less intimidating by using one of the following. (I've been heavily using R for ~3 years, and still find it more efficient to use one of these.)

• Rstudio (https://www.rstudio.com/)
• Rcommander (http://rcommander.com/)
• Rkward (https://rkward.kde.org/)

nice question and excellent comprehensive answer