Abstract:

In this work we present two computer applications, developed in Excel, specially designed for teaching statistics, as a tool, for those who need to apply it to analyse their data, to obtain results and draw correct conclusions, understanding how variations in data affect to the results and, ultimately, to the conclusions drawn from them.

Our goal is to design an interactive environment in Excel in which raw data, statistical tools, intermediate calculations, results and conclusions of the analyses coexist and interact in the same ecosystem, so that the intense and complex relationships between these become clear to the user/analyst, who can manipulate the data and the options of the statistical procedures appreciating, in real-time, the changes that occur in the analyses and the results, thereby acquiring a holistic vision and, therefore, a better understanding of the system under study. In this way, an attempt is made to give more prominence to the mentioned system than to the tools themselves.

In this work, we focus on statistics as a tool for those who will need to apply it, not to develop new methods or more precise or efficient methods, but to analyse data and draw conclusions. We think of users of statistical tools who are interested in obtaining correct conclusions from their data and in properly documenting these conclusions.

Of the innumerable existing statistical tools, we will focus on those related to a single continuous variable, either in a single population or in two independent populations. We will cover descriptive and inferential tools and, of the latter, we will consider both basic, parametric and non-parametric tools, as well as those based on resampling. Of the latter, we will focus on the bootstrap method.

In both applications, we have a previous interface for capturing data and selecting basic options. From this interface, once the user has specified the data origin and has selected the basic options, the application builds from scratch a complete spreadsheet that will contain the data itself, the intermediate calculations, the results (tables and graphics), as well as a basic wording of the conclusions.

The differential element of these applications is that, once the spreadsheet is automatically prepared, it is fully interactive, adapting to any change in the data and reflecting it immediately in the results and conclusions.

For the study in a single population, a descriptive study is performed: mean, variance, quartiles, skewness, histogram, and frequency polygon with the desired number of bins and with an assistant to choose the extremes, box plot, abnormal data detection, frequency table, normal probability plot. Standard parametric inference tools: parametric confidence intervals and unilateral and bilateral hypothesis contrast for mean, and the same for variance and standard deviation. Finally, we include advanced non-parametric inference tools: bootstrap confidence intervals for mean, variance and standard deviation.

Regarding the study in two independent populations, we begin with a descriptive and inferential study for each of the samples and then compare both samples with standard inferential tools (with and without homoscedasticity), with non-parametric inference tools (U of Mann-Withney) and with advanced non-parametric inferential tools (bootstrap). In each case, confidence intervals are included for the difference in means and hypothesis tests for equality of means and equality of variances.

Keywords:

Statistic tools, excel spreadsheet.