Scientific Program > Topic 7 > Session 7G >
 Presentation 7G2. A Data Analysis Tool that Organizes Analysis by Variable Types

 Presenter Rodney Carr (Australia) rodneyc@deakin.edu.au

 Presentation Abstract "What test do I use?". This is maybe the most common question asked by students learning statistics. This question, or ones like it, are caused by students being bewildered by the huge variety of analysis tools that are available. They seem to have missed the classes where their instructor told them that the type of analyses they can employ are determined to a very large extent by the number and type of variables involved. Probably essentially all teachers point this out one-way-or-another and essentially all textbooks also use the same idea to one-way-or-another organize the various topics. It is a very important guiding principle. In this talk we provide some evidence that one reason the principle is not recognized and used by many students (and hence the reason for the "What test do I use?" question arising) is that most of the standard statistical analysis software packages are not organized this way. We also demonstrate a package (XLStatistics) that is organized this way, according to the number and type of variables. XLStatistics is a set of Excel workbooks for data analysis that starts out by requiring users to identify the number and types of variables involved. Many (most?) textbooks start out this way, too, beginning with a discussion of the different types of variables. But later on many textbooks unfortunately switch to an emphasis on population parameters; XLStatistics emphasizes the "number and type of variables approach" all the way through. XLStatistics only requires that variables be classified into the broad categories "Numerical" and "Categorical". Finer classifications are, of course, sometimes important, but in XLStatistics the corresponding analyses are provided as choices later on in the analysis. Also, in cases where a variable can be treated as numerical or categorical (such as ordinal variables from a Likert-type scale or where dummy numerical variables are used for coding a categorical variable) the user can chose as they like. In all cases appropriate analyses are provided. These are actually contained in separate workbooks, but the user sees only an integrated package. Once a user has decided what type of variables they have, they simply highlight their data and select the corresponding XLStatistics workbook. Their data is then automatically entered into the workbook and the appropriate analyses are provided. At this stage the user usually has some choices to make, but the range of choices is quite limited and most students have no trouble making any necessary selections. XLStatistics provides most of the standard analyses. But in addition, XLStatistics also provides tools for "special" types of analysis. For example, even simple operations like switching the axes in a scatterplot (making the dependent variable go vertically) can be quite tricky in some packages - such operations have been built into XLStatistics and are accessed at the click of a button. (These operations are all standard things allowed by Excel, but as it should be for most cases, students do not need to know the technicalities of the package.) The evidence for the success of the approach comes from student evaluations from courses that use XLStatistics.