Statistical Software and Blind Users

created and maintained by Jonathan Godfrey

School of Mathematical and Computational Sciences, Massey University,

Palmerston North, New Zealand

Last updated: 17 December 2022

This page was created as a vehicle for showing which statistical software could be used by blind users who rely on screen reader software to have access to printed text. Most of the experiences listed here are my own, but I am grateful to other blind people who have shared their experiences with me.

Please contact Dr Jonathan Godfrey if you wish to ask questions about software not listed here that I might know something about or for more detail than is given here. Sharing your personal experiences with me might also help improve this page.

Formal publication of findings

In 2013, Theodor Loots and I started to collaborate together by getting much of the material that follows on this page submitted to a journal. Ultimately our work was published as a software review for the Journal of Statistical Software. This journal is open access so provides the best option for getting what we have to say out there.

Our review concentrated on four major statistical software options (R, SAS, SPSS, and Minitab) but also includes a list of criteria by which all statistical software can be judged. Take a read of the article by visiting the journal's page at JSS Volume 58, Software Review 1.

R

In its base form, R is perhaps unique in that it can be installed and used without any modification. There are ways to improve the experience of the blind user, such as using the terminal window instead of the GUI-based R console under some operating systems.

The use of commands as the standard way of operating means that the blind user works the same way as any sighted user. Note however that many R users are using new front ends (detailed below) to enhance their effectiveness. In general these are not useful for blind users.

I demonstrated use of R at the 2011 Summer University held in the Czech Republic. An R workshop facilitated by Bert van Landeghem was part of the 2012 Summer University. I returned to the Summer University in September 2013 held in Germany and then in Paris during July 2014.

My work on highlighting the accessibility of R for blind users was published in the June 2013 issue of the R Journal. A pdf can be downloaded from the link at the bottom of this page, or by visiting the journal's page copy of this 2013 R Journal article The full citation is:
Godfrey, A. Jonathan R. (2013) "Statistical Software from a Blind Person's Perspective: R is the Best, but we can make it better", The R Journal 5, pp73-79.
This article outlines some suggestions for making R an even better tool and highlights a number of other accessibility issues faced by blind R users.

One problem highlighted in this article was the loss of focus in R's terminal window under Windows 7 and Vista. The problem exists for users of every screen reader we've tested (JAWS, NVDA, WindowEyes, and Cobra). Thankfully, the problem went away in Windows 10. In January 2014, Dr Robert Erhardt (Assistant Professor of Mathematics at Wake Forest University) contacted me with a solution. His suggestion is to hit the ALT key when the focus gets lost. I've tested this with JAWS 14 and R3.0.2 running under Windows 7 and it worked for me. The editors of the R Journal accepted an addendum to the 2013 article which was printed in the June 2014 issue. Get the addendum to the 2013 R Journal article but please cite the 2013 article to help get its citation count higher.

I established a blind R Users Group (BlindRUG) email list in late January 2015 so that those of us using R can bounce ideas off one another. I intend to promote the updates to the BrailleR package on this list. To join the list, send a message to:
BlindRUG-request@nfbnet.org
with the single word:
subscribe
in the subject line. You will be sent a confirmation message to the address from which the message was sent within a few minutes. Simply reply to that confirmation message and wait for the initial welcome message with the standard list admin notes and a few initial pointers.

While I believe R is the best option for blind users, it is not perfect. There is still room to improve the user experience, especially for novice users. The BraileR add-on package is under development. More details on the BrailleR Project homepage. Last version used: 4.2.2

R Commander (Rcmdr for short)

R Commander is a set of menus and dialogue boxes that produce standard and nonstandard R code to run analyses. The dialogue boxes are not screen reader friendly. The code that is generated is as useful as base R code.

Any student needing to work alongside sighted students using Rcmdr could generate the same results using the code by typing it out for themselves in exactly the same way as they would standard R code. Having access to this extra code does require the use of the "library" or "require" command then closure of the R Commander windows.

Last version tested: 1.9-2

RStudio

RStudio is a development tool for combining R commands, output, and graphs with regular text for creating a blended document. It is not accessible for screen reader users (Windows and Linux have been tested) and there is no parallel way of working using all of the RStudio practices; alternative modes of operation can be used such as use of R markdown within base R functionality. I use R markdown documents as my primary means of working, and I don't use RStudio.

RStudio is putting effort into its accessibility

Last version tested: 1.3

Tinn-R

I've had a quick look at this development tool/text editor for linking R work together and managing it all. It is accessible but I suspect that the learning curve for Tinn-R on top of the R learning curve might make it a poor option for the novice user. If I was to choose between it and using R markdown, then markdown is the way to go.

Read more about it via the Tinn-R homepage where a link to its download can be found.

My only major concern is that Tinn-R seems to have taken over a few things about my use of R in the normal fashion. When I start the R terminal it automatically loads various add-on packages that Tinn-R wants. Personally I don't like this so Tinn-R is no longer resident on my machine. This is because I need to have R running in the same way as the students I teach will have it running. So sorry, please don't ask me about the use of Tinn-R. I'm not going to be able to help much, and I'm going to recommend that you not use it as well.

Last version tested: 2.3.7.3

Sweave

I used Sweave extensively for document preparation purposes because I could read the output from my R commands as well as create my LaTeX document. When writing up the .Rnw file, I do not get tangled up in R output etc. I avoid distorting the output from R that will be seen by the sighted reader. It also avoids the need to cut and paste like sighted people have to do when working in other ways.

I've successfully used Sweave documents within the Texnik Center editor for LaTeX (Windows only). I'd like to learn more about similar ways of working in other operating systems and/or environments. The way to get a file working with Texnik Center and Sweave is to tell Texnik Center how to process an Rnw file. The settings I incorporated into my Texnik Center settings are in the file Sweave.tco which can be downloaded using the link at the bottom of this page. If you do download this file you will need to look at it carefully, to ensure your system's arrangement is similar enough to mine for the settings to be successfully imported into Texnik Center on your machine.

Having said that, I now only use Sweave with legacy documents because I prefer to use R markdown instead. Unless you have some superb skills in LaTeX and want to retain the flash extras that LaTeX offers over R markdown, I'd recommend avoiding Sweave. Novices shoudl definitely use R markdown instead of Sweave.

Last version used: R 4.1.1

R markdown documents: the knitr package

The knitr package is available from CRAN using:
install.packages('knitr', dependencies = TRUE)
If using this package in conujunction with LaTeX files, it is a step up on basic Sweave in that the power to do things is increased, but so is the relative complexity.

I've started using knitr in conjunction with the documents I started under Sweave. There were some changes required of course but I'm happy enough with the results. Moving to knitr is probably a good long-term solution but I would recommend that blind users get to grips with Sweave first.

Having said that, the R markdown files that are the bread and butter of the knitr package are really cool. This is the way to go for all R users (that don't need the full flexibility of LaTeX) in my book. The resulting output files can be really accessible to the blind reader because we can have our html while the sighted have their pdf, all from the same source file!!!

Last version used: knitr 1.34 under R 4.1.1

Minitab

Version 14 used a different worksheet for data which was not accessible. Another issue is that some dialogue boxes new to this version were not accessible. This includes the design of experiments and graph creation dialogues.

Versions 15 and 16 have been altered to improve the visual experience of the sighted user. Menu items that have no relevance at the specific time that a user pulls up that menu are greyed out. This causes screen readers grief and the wrong item is read aloud in the majority of situations. I have used both of these versions (unwillingly) to create material similar to that created by my students, but my ability to do this is heavily reliant on my memory of the menus and dialogue boxes built up from previous versions of Minitab. A short audio recording showing some of the Minitab problems, is at the bottom of this page for you to download. It uses NVDA but my previous experiments with JAWS gave similar results.

I tested version 17 using JAWS and NVDA and found the menu items were all inaccessible. I also found it impossible to set the options to what was needed for my successful interaction without assistance from a sighted helper; this includes enabling the command language which is essential for use because the menus are inaccessible. I also found it very difficult to find out how to get the commands for specific actions to be generated so that I could then save them for later use.

Use of the command language of Minitab does remain an option. I started working this way in version 8.2 (for DOS) in 1995 while my sighted classmates were using a windows version (10.1 I think). My reason for using 8.2 was not a Minitab issue but was due to the inability of blind people to successfully work with the Windows operating system prior to Windows 95.

The command syntax is not as intuitative as other languages and it is difficult to find people who are comfortable using the command language today. I have found examples of dialogues that are not able to be replicated by commands, and not all commands are covered within Minitab's own documentation.

Last version tested: 17

Minitab Express

Hot keys do exist for opening files and pulling down the right menu, but the ribbon-style of finding the right dialogue was frustrating. On many occasions, the use of cursor keys to move around the ribbon led to confusing feedback and did not always lead to the right ribbon element being found. The dialogue boxes were quite clean compared to the ones found in the standard Minitab version.

I could not tell what was in each cell of the data worksheet; I could not tell what was in the output window as all output was in the form of a png graphic; I found no way of accessing the functionality of menus and dialogue boxes using commands as I would have used in other Minitab versions.

In summary, I was hugely disappointed in this product which might have solved the problems I've identified with other Minitab products.

As an aside, this product is not meeting with uniform support from my colleagues either. They complain about the inability to complete some tasks that are fairly standard features. They identify the problem of editing a graph for making it publication standard in particular.

SPSS, also known as PASW

SPSS has been used by blind people with some moderate success. In versions since (perhaps) 15, SPSS has made extensive use of Java for presentation of analyses. The Java Access Bridge is therefore required to make full use of SPSS. This is a separate installation process for versions of SPSS up to v21, but it is hoped that v22 will more fully automate the process for blind users.

While you might persist with the menus in SPSS, you will almost certainly benefit from using the command language that lurks behind the menus. In the early days of SPSS, typing commands was the norm, but finding resources that are current, or humans that are comfortable using the syntax of SPSS may be difficult. The commands can be printed to the syntax window from any dialogue boxes by clicking the Paste button. Note that this does not submit the code for processing in v21.

Last version tested: 21

Stata

Stata can be used by blind users. I attended the 2011 Stata workshop facilitated by Bert van Landeghem at. See details of his other Summer University workshops at 2010 Stata workshop and 2012 Stata workshop, but note that this time Bert mentions R as well as Stata.

A copy of all commands and associated output, or the commands alone, can be saved in log files. These log files need to be saved as plain text so that you can read them using a browser of your choosing. The reason for opening a plain text file in a browser is that the log file will be updated as you issue new commands in Stata; pressing f5 to refresh the browser will get the added content from the log file on screen for you to review. This use of a browser is laborious. Use of text files that contain the commands you want to issue is a better idea. These are called "do" files by Stata.

Access to the help documentation for Stata commands is best obtained by searching the internet because the built-in help does not work with a screen reader. Some menus are accessible, while others are less so; I find this inconsistency irritating to say the least. When these dialogues are accessible the underlying commands that these dialogues generate can be copied to the clipboard for insertion into a "do" file.

Typing commands at the command prompt is an option worth considering for simple jobs; it can be made simpler by learning that a command can be abbreviated to a few letters if it is then unambiguous, and tab completion can help enter the names of variables. For example, if my data set includes a variable called "MaleLifeExp" and there are no other variables starting with an M in my dataset, the command "summarize MaleLifeExp" can be shortened by typing "su Ma" and hitting the tab key followed by the Enter key. You should get into the habit of saving your commands using a log file. This is much easier than trying to save the work you've already done. The necessary command will be of the form: "cmdlog using Today.txt"

When you start to use the "do" file way of working, you will soon find that the highlighting of text will not be read aloud by your screen reader. These files are just plain text files though so perhaps the most pragmatic thing to do is to open the file using your favourite text editor software.

Last version tested: 14

SAS

I initially tested SAS on a network computer using the NVDA screen reader (free) because I didn't have direct access to SAS from my own machines. Access under the tested version (9.3) was similar to that used way back when I was a postgraduate student using version 6.2. SAS programs are just text files so are fully accessible. SAS output files can be saved as plain text and graphic files or in html format. Both are readable.

I've worked with Theodor Loots (University of Pretoria, South Africa) to make use of his far superior knowledge about the best ways of getting SAS to work for blind users. Read the article using the link at the top of this page.

SAS help documentation is freely available on internet pages. These often have some theoretical guidance mixed in with the necessary command syntax documentation. This material is very accessible for blind users.

Perhaps the only drawback to using SAS for blind users is the one often heard from sighted statisticians - too much output to sift through. It's true, SAS does generate a lot of text to read through, but if this is the software used by your class/workmates, then using it is to be preferred over using an alternative. Reviewing the output in html reduces this problem a little, and efficiency will improve with familiarity.

I have not tested the enterprise guide or any other add-on package/software at this time. I did look at Enterprise Miner around 2003, but that required mouse clicks that i could not emulate via the command language. This is something to watch out for but I do not express any opinion about the current state of play for this add-on software.

Last version tested: 9.3

MPlus

I obtained the following summary of Arielle Silverman's experiences of MPlus via email lists I follow. Please contact Arielle if you wish to ask questions about this software.

In Arielle's words:

Just wanted to share that I learned to use MPlus (software for analyzing structural equation models, and I think it can also do more traditional regression) and I found it to be extremely accessible with JAWS, even more so than SAS. I could easily enter and edit syntax, run it using a simple menu-based command, and then read the output directly in the output window and move between input and output. I also never had to use the JAWS cursor! Though the syntax is fairly easy to learn, it is also possible to run most MPlus operations using a menu-based language generator, which is also extremely accessible (much more so, I think, than SPSS).
Anyway, I would recommend MPlus to any blind folks looking to do regression-based analyses. I used a free demo version, which allows input of up to six dependent variables and up to two independent variables. This is fine for simple regression, and for structural models involving only one or two exogenous (independent) variables, but it might not work for multiple regressions with more than two predictors. However, , there may also be temporary licenses available that are more forgiving.
I had someone help me a little bit with syntax, but many syntax examples are available online, and the command structure is very intuitive. For example, to regress Y on X you would simply write (after specifying your datafile and variable names), "model: y on x").

(Arielle's words end)

Genstat

My only foray into Genstat was a complete failure from an access point of view. No access to anything useful was to be had - input, data, and output were all inaccessible. The accessibility of the menus was irrelevant as a consequence.

Last version tested: Genstat for Schools (2010)

JMP

In November 2013 I had a chance to see some of my colleagues using JMP in a workshop setting. The continual spawning of subwindows that can occur is probably a major drawback for screen reading software.

Last version tested: 10

Statistica

My only foray into Statistica was a complete failure from an access point of view. As I recall, the menus were not even accessible in this software.

Last version tested:

Excel

I put Excel in this list as I know there are people using it to complete statistical tasks. I do not like relying on it as I don't consider it to be statistical software but it does have its uses. The problem is that while Excel is fairly accessible, the range of features from a statistical standpoint is fairly limited.

Add-on packages are commonly used to bridge this gap. Access to the functionality of add-on packages should not be muddled with access to Excel. Thankfully, the add-on packages for Excel that are used in educational settings are commonly distributed with a textbook and are therefore free. I can only advise that testing be done before anyone relies on access to the functionality of add-on packages. There are just too many add-on packages to mention them all here but one does get special mention.

At the 2014 Summer University, Anthony Reas showed how the add-on known as XLSTAT can be used by blind people. I didn't attend his workshop because I was hosting a workshop of my own, but feedback was good and I've asked him for his version. In Anthony's words:

XLSTAT is one of the more comprehensive add-on packages that adds statistical power to Microsoft Excel in regards to both accessibility and feature set. Being integrated into Microsoft Office's spreadsheet software, most data-entering and preparation procedures can be handled with the standard Excel options.

When the add-on is activated, XLSTAT creates a new Excel ribbon, from which all available analysis tools can be accessed using a menu-like structure, dividing into categories and subcategories. Once an analysis tool is selected, a dialog box appears that lets you select data, typically by entering data-regions into edit fields, and selecting the desired options and outputs for your analysis using simple check boxes and drop down lists. In addition, each dialog contains a button that leads you to the relevant HTML help page which is readable and understandable without any further issue.

Overall, these dialog boxes are rather accessible other than one major drawback. In more recent versions of Microsoft Office, data-regions entered in data entry fields are no longer announced by screen readers and cannot be reviewed by the arrow keys without a review cursor, as using them will change the actual data selection. However, the amount of text that has to be entered in those fields is rather small, which makes this limitation easy to overcome. Also, XLSTAT output can be classified as accessible to a large extent, for its textual part is generated as a standard Excel table, which can be read, manipulated, and saved in the same way as any other Excel data. Additionally, the graphical output can be saved and pasted into another application. Still, whether these graphs could be embossed to a good quality tactile form remains an open question for now. The base package of XLSTAT comes with enough features to suffice the usual needs of a social or behavioral science student partaking in an introductory statistics course or working on smaller research projects. These features include regression analysis and ANOVA, hypothesis testing, factor analysis, etc. Furthermore, a wide array of feature extensions available from the XLSTAT website are able to add functionality for more specialized designs. While a base student XLSTAT license costs an annual 50 dollars, and the base feature set may not be as rich as is present in some other statistics software, its rather high accessibility level and its easy-to-use dialog based interface could make it into a viable working solution for students and professionals who are less comfortable with command-driven tools. Especially those who are already familiar with Excel and will, as a result, experience almost no learning curve, could benefit most from this product.

(End of Anthony's words).

Please double click the following files to open/download

File Name:

Godfrey2013RJournal.pdf

MinitabWoes.mp3

RConsoleUnderWin7.mp3

RTermUnderWin7.mp3

RUnderWinXP.mp3

Sweave.tco