Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Jason Brinkley, an Assistant Professor of Biostatistics at East Carolina University, recently recorded a set of videos for JMP titled "Moving from SPSS to JMP," which address the differences between the two software packages. After taping, Jason answered some questions that dig a little deeper into these differences.
What differences will be immediately apparent when a SPSS user begins working in JMP?
The most immediate difference is in the ways each software chooses to show the data. Both SPSS and JMP do show data in the "spreadsheet" style -- with variables as columns and rows as individuals. However, SPSS has a variable view, which shows other information about the columns of the data, whereas JMP puts much of that same information to the left of the data.
Are there inherent differences in the two systems, or is it just a difference in functionality?
There are inherent differences, but they stem more from the different approaches to analyzing data. The SPSS archetype can be more analogous to an experimenter, mostly focused on preplanning and designing exactly which analyses should be run and then selecting only those options. The JMP archetype, by contrast, involves more data exploration and is designed for users to be able to work with data more on the fly. That isn’t to say that SPSS can’t explore data and JMP isn’t great for preplanned analyses; we are just considering what seems to be their native environment.
Visualization is a part of each product, but is the process of creating visuals the same for both?
The two platforms are most similar in this area, each allowing for customized graphics, but it is where we see the biggest differences between the archetypes. JMP allows users to see the graphs they want to create as they are creating them, allowing more flexibility and imagination to go into creating visuals.
Does SPSS or JMP require certain knowledge of statistics to really appreciate what they can do?
I always tell my students that these software packages are only limited by how much statistics you know and understand. I think both software packages can be used with very little statistical training, but the perceived value may be different between knowledge groups. Features such as table creation and data visualization come easier to individuals with limited statistics training, in which case the Graph Builder and Tabulate features in JMP may be more intuitive.
What would you say are the strengths and weaknesses of each?
JMP is strong at data exploration and visualization, and at being able to quickly and easily manipulate as well as exclude both data points and unused output. SPSS is strong in the different analytic options available in the software and the ease with which one can get very specific output quickly. The idea of customizing output in JMP causes certain analyses to be harder to find or find in nonintuitive places. For example, JMP does not have as many modeling options as SPSS, and repeated measures is not in an intuitive place for researchers in health or social sciences to quickly find it. SPSS does not have as many options for understanding particular nuances of or unusual observations in a data set.
What’s the best way to see or understand the differences between JMP and SPSS?
Let’s take an example that may illustrate the strengths and weaknesses of each. Suppose I conduct an experiment where I follow 100 individuals across four time points as they work through some weight loss regimen. Suppose I randomize the 100 individuals to five trainers (so 20 to each). Let’s say that two of the trainers are much more experienced and adept at getting their clients to lose weight. Suppose I also design a survey to give to the subjects at the end of the program to get feedback on their opinions of the weight loss program. JMP will likely perform strongly on the visualization, and users will be able to quickly see if the individuals in the program did in fact lose weight across the four time points. JMP will also point out early on in data visuals whether one or more of the trainers had subjects that outperformed the other trainers. Suppose a researcher wanted to bypass the visuals and exploration and go straight to some sort of repeated measures ANOVA or longitudinal modeling; in that case, the options are easy to find in SPSS, and obtaining output is straightforward. Next, if the researchers wanted to look at the survey data, JMP will come on strong as users can understand broad survey results quickly and explore differences across trainers with the Tabulate and categorical platforms. Should the researchers be concerned about their survey’s reliability and consistency, the options for obtaining that output will be easier to find in SPSS.
Any additional other comments?
No software is perfectly suited for all analyses, and the archetypes of each can make it more ready to handle certain problems than others. There is a balance between the ability to perform powerful analyses, options for customizations and ease of use. The example described above can be easily worked out by any user with a small amount of experience in either software, but there are aspects that will be more challenging for new or transitioning users. The university I work at is fortunate to have licenses for each program, and I know many SPSS users who love JMP and are making the transition because they feel as though they understand their data better in JMP. Visualizations tell a powerful story in a way that numbers and analytics can’t. But many researchers know that visuals can also lead people astray, and our brains can be fooled; there is still a need for analytics to support what we "see" in the data. SPSS users have to retrain themselves to the JMP archetype, which may mean that they have to think about where the options to do certain analyses will be located in the software. There is no nonparametric statistics button in JMP, and users have to put their problem in a certain context to determine where they will find the options for a Wilcoxon test (hint: it’s the nonparametric alternative to the t-test so look for it there).
Thanks to Jason for his insightful analysis. For more information on this subject, check out the Moving from SPSS to JMP Information Kit on our website. The kit contains a white paper and Jason's video presentations with slides and a data set.