I recently started teaching statistics and was surprised to find that the vast majority of resources that were available for students were "static" resources. In other words, students had tables for computing probabilities from z-scores, or programs that given a set of data, computed the mean. What I wanted was a dynamic tool that allowed students to immediately see the impact of their data. For example, what is the impact of deleting a value from a sample on the quartiles, mean, and mode? Existing tools allowed them to compute these values. The tool I wanted would show them visually what happenned immediately in response to changes. As another example, we teach students about hypothesis testing and use either P-Values or critical values to evaluate the outcome. I wanted a tool that showed graphically, immediately, what the impact was of changing the signficance level or the claim on the result of the test. This tool is a first attempt to do that while I continue my search for an existing tool that has similar functionality.
This tool is not intended to replace a complete statistics package such as Minitab. In particular, I have made no effort to include all of the possible detailed values that could be computed on a data set. This tool is meant to be an educational package that allows introductory statistics students to get an intuition for statistics. Once that intuition is establed, the results of statistics packages will be more meaningful.
Currently, dStat has interative support for four branches of statistics:
- Descriptive Statistics. dStat allows you to enter a 1 dimensional set of values and watch as various descriptive statistics are recomputed as the data set changes. These measures include Mean, Median, Mode, Variance, Standard Deviation, Minimum, First Quartile, Third Quartile, Maximum, Interquartile Range, and Range. Also shown as the dataset changes are a box plot, histogram, and dotplot, with support for overlaying some of the descriptive statistics on these displays.
- Probability Distributions. dStat supports both discrete and continuous probability distributions. For discrete probability distributions, binomial properties can be entered, or a custom probability distribution can be provided. For continuous probability distributions, both Normal and Student T distributions are available. With each distribution, students can click in the distribution and see the resulting test statistic (z-Score), left probability, right probability, and unnormalized value that results. In addition, each of these values can be changed and the rest will update automatically to accomodate the new value. So, for example, if students enter a new left probability, the right probability, test statistic, and unnormalized values are all immedately recomputed to reflect the change.
- Hypothesis Testing. dStat currently has Proportion and Mean hypothesis testing support. Given the claim, significance level, and summary statistics describing the sample, dStat will display the situation graphically showing where the critical value and test statistics lay, as well as the relevant probabilities.
- Regression Testing. dStat allows you to enter a two dimensional table. The least squares fit for the data is computed dynamically as the data is entered. The Pearson coefficient, regression coefficient, slope and intercept are displayed as each new piece of data is entered.
Presently, an installation package is available for windows (see below) as well as android (see the Google Play Store). The underlying software is written to support OSX as well. At some point in the future if there is need I may make an installation image for OSX, but first, I would need an appropriate computer to test it on.
Note: This program is currently under development.
The current Windows installation image can be found here: dStatInstaller. This will create a link to the executable on your desktop.
I also have a portable installation zip file that can be found here: dStat.zip. Extract this zip file. Then click on dStat.exe to run the program. This does not create links or install the program. You have to click on dStat.exe each time you want to run it.