**3.2.3 Statistical software**

We received 48 questionnaires related to the use of the data visualisation functions of statistical software in the social sciences. From these results it is clear, however, that statistical packages are primarily being used to store, manipulate and quantitatively analyse data. The graphics functions of these software packages were not widely used.

As the graph shows, SPSS was by far the most popular statistics package in use in the social sciences. In total around 60 per cent of users of this group of software were using it. Minitab was the next post popular statistical package with around 15 per cent usage. SAS was next most popular followed by Statview. For other statistical software only one questionnaire reply was received in each case.

Statistical software was widely used throughout the social sciences. Unlike visualisation tools such as GIS software it was not predominantly used in any one subject area. We received most reports on the use of statistical software from geography, social statistics and economics. We received one questionnaire each from a number of other subject areas. Interestingly, and no doubt a product of the limited penetration of this survey, we received no reports at all on the use of statistical software in history.

In detail responses were received for the 13 different packages.

**ExperStat **

We received one questionnaire referring to ExperStat used on a Macintosh OS 7 or 8.

ExperStat was used in teaching as a straightforward introduction to ANOVA, but the software also has the capability to carry out other statistical tests. The software's particular strengths were listed as

- Simple data entry
- Good documentation
- Clear tables and graphs
- and, overall ExperStat's simplicity

**LIMDEP**

We received just one brief report on the use of LIMDEP, an econometric package in this instance used in research. The software was described as being fairly easy to use and very useful. Its main value was in the increased productivity it afforded.

**Matlab **

One respondee discussed the use of MATLAB in teaching. The software was described as being neither particularly easy nor difficult to use. It was viewed in high regard however.

Specifically the software was used for

- Matrix manipulation
- Use of the language script to customise statistical tests
- To visualise data
- Analyse data with customised procedures
- To produce graphics

- Ease of customisation
- Informative error messages
- The ability to handle large matrices

In terms of software development an interactive graphic editor would be of value.

**Microfit **

We received one report on the use of Microfit for econometric estimation in teaching and research. Microfit was described as both being very easy to use and very useful.

The software's main advantage was described as its menu driven interface. This was also listed as its main disadvantage!

The most useful software development would be the ability to write macros to be inserted into the current user interface need to keep up with statistical developments.

**Minitab **

Minitab was the second most popular statistics package in use in the social sciences. It was used both in teaching and research although some users regarded it as a more limited research tool. No one found Minitab difficult to use and it was generally regarded as being either 'very easy' or 'easy' to use.

Minitab was used for:

- Statistical analysis including descriptive statistics, tests of association, tests of difference, control charts, tables etc
- Result plotting such as histograms, boxplots, regression plots, scattergrams, bar charts etc
- Introductory Data Analysis year 1 Social Science

- The flexibility of command line syntax allowing instructions to be written in
- English.
- The menu-driven interface on later versions
- User Friendly
- Good help facilities
- Does not distract from statistics by over fussy computing
- Has all main techniques available
- Very basic quick draw graphics for quick look in analysis
- Good graphics and output - although other users suggested that the graphics were poor!
- Less steep initial learning curve than many statistics packages
- Well adapted for teaching use, but good enough in recent versions for many research purposes
- Powerful macro language for writing purpose built special analyses

- Inbuilt routines are insufficient to allow this to be used as an advanced tool for applied research
- Poor graphics - although later releases have steadily improved Minitab's graphics facilities. One user wrote 'Has come a long way in improving its graphics, which were a major weakness.' However, another wrote 'lousy graphical capabilities'. I great deal depends on the version of Minitab being used.
- Cannot easily read data from other packages
- Spreadsheet is not as easily manipulated as in other packages such as Excel
- Tabular output is not as good as SPSS - needs the concept of value labels for variables
- Limited coverage of econometric methods
- No built in optimisation for Holt Winters' forecasting
- Poor 'help' in some instances
- 'The students always forget how to enable the command language - then find the multiple windows confusing.'

- Development of a library of off-the-peg subroutines
- Better and more flexible spreadsheet functions
- Better logic and transfer of control functions for batch programs
- Dynamically linking graphics and data
- More econometric facilities built in
- Better help pages

**SAS **

Surprisingly we received only one questionnaire describing the use of SAS. In this instance SAS was being used in both teaching and research.

Whilst SAS was regarded as being very useful, it was also described as being very difficult to use.

Specifically SAS was described as being flexible. The graphics module is part of very powerful statistical and data transformation package. Its main disadvantage was its difficulty of use - SAS is not a package for the novice user.

The most significant software development would be an easier to use front-end for charts etc. along the lines of what is available in competitor packages such as SPSS.

**Schema**

We received one report on the use of Schema in research. The user felt that the software was both easy to use and very useful. Schema was used for data analysis, data manipulation and data representation.

The software was particularly valuable because of its speed and ease of use. No drawbacks were identified.

**SECOS**

SECOS, produced by Statistics for Education, is available for both DOS and Windows platforms. It offers a front-end to statistical datasets allowing data manipulation, analysis, graphing and mapping. The software is particularly useful in teaching quantitative techniques to students.

We received one report on the use of SECOS for teaching levels I and II and Masters students in Sociology and History.

SECOS was considered to be very valuable and fairly easy to use - 'Sociologists can use it anyone can!!!'

SECOS was used for:

- Mapping
- Graphics
- Statistical Testing
- Extracting Data
- Analysis of specific pre-prepared datasets such census statistics and social and European trends

- Easier than going to the library to look for statistics
- The mapping interests enumerate students
- The statistical tools are easy to use

**SigmaStat **

This statistical package was not widely used in the social sciences in the United Kingdom. In fact only one questionnaire made reference to the software from a respondee in the United States. SigmaStat requires a 486 IBM PC running Windows95.

The respondee used the software for both teaching and research. Specifically the software was used for descriptive statistics, chi-squared analysis, regression analysis and t and z tests.

The software was considered to be easy to use and within the constraints of the software very useful - 'This is a good starter package. Not very powerful, and not all that flexible, but simple and instructive.'

It was considered to be particularly valuable as a teaching tool. The software
includes an *analysis wizard*, which helps to determine the best analysis
to run. Following statistical tests it gives an indication of significance
levels where appropriate.

SigmaStat is limited in that data inputting is not as straightforward as with some other packages. It will not allow alpha data to be coded and the range of statistics is limited compared to some other software.

The software could be improved by upgrading the analysis wizard and the interface more generally.

**SPLUS **

Another statistics/graphics package used infrequently in the social sciences. The software was used in research and for the creation of diagrams for papers. It was considered to be very useful but not particularly easy to use.

Particular uses/advantages of SPlus included the good 3D graphics and the non-parametric output. However, some programming using the SPlus language is usually required. The language has become easier to use over time. Also it is often hard to get things exactly right re: margin sizes etc.

**SPSS **

SPSS was by far the most frequently used statistical package in the social sciences. In total 27 questionnaires were received describing its use. SPSS was used in both teaching and research. Some respondents used SPSS solely in teaching, others solely in research and others in both activities. Generally SPSS was found to be 'easy' or 'very easy' to use. A small number of respondents found the software more difficult - particularly when students were being taught to use it. Virtually everyone found SPSS to be a 'very useful' tool.

The software was used in a range of applications including specifically:

- Charting Russian voting patterns and socio-economic characteristics
- A class developed with assistance from the US Department of Education called New York city's Lower East Side, 1880-1920.
- Teaching postgraduate masters students and also in applied research to study survey data.
- Teaching at undergraduate and postgraduate levels. Mainly in Quantitative Data Analysis and statistics based modules. Also for individual dissertations.
- Analysis of data from the Bangor Longitudinal Study of Ageing and Older People in Europe's Rural Areas.
- Analysis of questionnaire data in teaching and research.

- Data Entry
- Recoding and transforming data
- Data management
- Preparing large and complex data
- Explore data
- Large-scale analysis
- Visualise numerical data
- Statistics
- Graphing
- Batch file Running

- Ease of use particularly with the Windows releases. Far easier to work with than DOS-based statistical applications.
- Very user friendly
- Covers most routine analyses
- Presentation quality tables and graphics. One respondent wrote 'I no longer have any need for Harvard Graphics, for instance, which I used to use.' Another wrote 'It took me a few weeks to figure out the nuances of making the charts, but once I read through the instructions, I figured out how to customise the charts and now find the software very efficient and useful.'
- University site license enables all to use of SPSS with no cost at the Departmental level rather than individual researcher/lecturers having to make a case for purchasing unusual software.

- Some criticism of the graphics facilities including problems in positioning graphics precisely on the pages, difficulties in editing graphs, limited graphical capabilities in early releases.
- Need to have some of the newer graphs - dot plots, mosaic plots, comparative bar charts with central base line
- The graphic editor is not very versatile.
- Automatic scaling: makes it difficult to present different charts using the same scale for comparisons.
- Need for fast processor for work with charts
- Too easy? Inexperienced users may be unknowingly applying the wrong or inadequate methods
- Can be off-putting to the beginner - the learning curve is still somewhat steep for some undergraduates.
- Lacks many regression analysis techniques.
- Difficult to edit output.
- Problems with importing and exporting data in a variety of formats.
- Poor error messages on occasion - not always easy to associate with correction needed to the code.
- Limited functionality for complex analysis
- In batch mode one error halts the whole file running.
- Can be slow at times if not using the latest hardware
- Difficulties running SPSS across the networks in relation to the large temporary datafiles it creates.
- You need knowledge of statistical terminology to operate the software and some of the language is specific to SPSS.

- More flexible editing of graphs including the ability to click and drag graph elements.
- Enhanced graphical capability would be an advantage to many
- More graph types and specifically better Stem and leaf charts
- Improved export facilities. One respondent wrote 'Particularly I would like to be able to export .spo graphics, maintaining the appearance that they have in SPSS'
- Made simpler still for undergraduates
- Manuals could be much more user friendly
- Improvements to syntax window
- Easier repeat procedures for analyses are required. Maybe some macro or learn feature, so that new data can easily be processed quickly using the same sequence of analysis commands applied to previous data sets.

'I prefer to use other packages for my research. I think my students find it quite difficult to use.'

'I provide the opportunity for students to obtain their own data rather than use that provided by the manufacturer. This makes learning personal and meaningful.'

'After an initial introduction to the system if becomes quite routine and reasonably easy to use.'

'It is sometimes difficult to get undergraduates into the package quickly although the later versions are a big improvement on earlier versions.'

'Once students understand the way that SPSS operates then they find it easy to use the software. The only problem that I can see is that it can be too easy for students to undertake statistical tests without understanding what and why they are doing it!'

'SPSS is far superior to SAS for the PC, although it has weaknesses in graphic output and some other basic operations needed in statistical analysis.'

'Still believe SPSS hasn't moved to true windows version. The 7.1 is a move in the right direction, I welcome some of the new features, interaction diagrams, some post hoc testing. The graphics environment is still slow and basic.'

**Statview **

Another statistics package rarely used based on our survey results. Statview was used primarily for generating statistical output. The software was considered to be 'almost idiot-proof' with easy output - the reason given for not using SPSS. The graphics was described as useful but the range of analyses were limited. The software would be improved by increasing this range.

**XLISP-STAT**

Good reliable software but a steep learning curve. Full programmability in a real programming language makes the system flexible, but a little harder to use than most.

Specifically the software was used for:

- Interactive, linked graphs
- Simulation abilities
- Basic statistics (descriptive)
- Advanced statistics (i.e. SLIM)
- General programming (i.e. coding )

- Freeware
- Programmability
- Flexibility
- Interactive graphs for exploratory data analysis
- Built in statistics and functions

The software could be improved through the introduction of some fairly specialised statistical techniques, such as spline smoothing - which may be used in conjunction with visualisation.

Graphics Multimedia Virtual Environments Visualisation Contents