Yes, I admit, partly another way to push the R project (!) but no, really, this is here because I think it helps to think that its development shaped research and outcome measurement. One sociological theory, Actor Network Theory (ANT) explores how sociopolitical changes are shaped by actors but always within networks and suggests that actors aren’t always humans, they can bugs (coronavirii?) or technologies (Newcomen’s steam engine, Watt’s addition of the external condenser, the car, the computer). My own view is that the evolution of statistical software has shaped our fields.
In the early 20th Century computers were people doing longhand calculations or more often, calculations assisted by simple adding machines or slide rules. (At school in the late 1960s I loved our linear slide rules but I was proud to have inherited a cylindrical slide rule, I think from my paternal grandfather, he and it long gone sadly.) There’s a whole sociological side of those networks and their gender (and colour in the USA) that’s emerged in the last decade or two. Then the computer as we know it changed all that: it ran software (an undersung cadre, mostly of women, moved from being the computers to programming them).
Now statistical analyses like a factor analysis, that would have taken a research working alone perhaps a week to complete, could be done in an hour, now that level of factor analysis even with many variables and perhaps tens of thousands of participants’ data completes on a laptop so fast you hardly notice a pause. However, the analyses still needed to be programmed and in a sensible language, say FORTRAN or later C, or C++, this was still the province of very skilled individuals. The game changer was the emergence of packages of statistical software. Since the early 1980s I have used: BMDP, SPSS, rolled my own (trivial one in DB-II), GLIM (very briefly), Genstat (even less), Systat (briefly), SAS and SAS/IML, S-PLUS and, finally, R. I’ve also used a few systems specific to particular tasks such the ones dedicated to factor analysis and more general “latent variable modelling” including LISREL and AMOS.
About 15 years ago I decided I would be more efficient if I stuck to one system so I haven’t used statistica or M-plus, respectively probably the most used recent general and specific statistical software.
Each package had particular strengths and weaknesses and mapped to particular application area. If you follow the links to some of the extant systems above you can see the mappings and commercial realities of the systems. For most of the last forty to fifty years SPSS has been the dominant package for psychology and mental health. I suspect its (often excellent) manuals all written my Marija M. Norusis in the early days, and what it did and didn’t do, shaped much of what quantitative research happened.
Most statistical software until the emergence of SAS/IML, S-plus and R was a collection of packages with specific tasks. To some extent that’s still true of all statistical software systems but what has changed is that some still essentially present a menu of what you can and can’t do (notably SPSS, though, yes, it is possible to get beyond that way of using it, it’s just difficult). These others still have specific functions and packages of functions but they don’t work in a “pick from the menu” fashion, they ask you to tell them what to do and to elaborate that where they may not offer you an “on the menu” option. This is potentially transformative allowing us to do things beyond what the system’s programmers created for us.
The other transformative thing about the R project is that it’s open source: entirely free of licence fees (which are steep for commercial systems). Try to transfer to it if you haven’t, teach it to your students if you teach!
Try also #
Use of statistical software is implied in Chapter 8 and you can’t analyse service level data without it. In fact unless your client numbers are tiny even single handed practitioners will probably need statistical software, or someone who can use that software to analyse their own data. Though …
Online resources #
First created 17.viii.23.