R-INSTAT INTRODUCTORY

TUTORIAL 1

BY: ROGER STERN, DANNY PARSONS, JAMES MUSYOKA , DAVID STERN AND BERYL JOHNS
November 2022
R-INSTAT INTRODUCTORY TUTORIAL PART 1

CONTENTS

Contents 1
1. Introduction 2
2. Exploring R-Instat 3
Starting 3
A first task – Importing data from the library 3
Some graphs 6
Some Summaries 10
A more ambitious analysis 11
Correlations 14
Outliers 16
3. Reflections 18
4. Next steps 20
5. Feedback and reporting bugs 20

1
R-INSTAT INTRODUCTORY TUTORIAL PART 1

1. INTRODUCTION

Welcome to this R-Instat introductory tutorial. R-Instat is a free, menu driven

statistics software powered by R. It is designed to exploit the power of the R statistical
system, while being as easy to use as other traditional point and click statistics packages.

R-Instat is developed under the African Data Initiative (ADI), a collaborative project to
support improved statistics and data literacy across Africa and beyond. The overall aim of the
African Data Initiative project stretches beyond producing this software, however R-Instat is
an important step in achieving change.

The original target audiences for R-Instat were described in the crowd funding
campaign which launched the development. We claimed there was a need for statistics
software that is easy to use, free and open source and that encourages good statistical
practices.

The "Instat" in "R-Instat" refers to a simple statistics package first developed in the
1980s with similar aims and target audiences as R-Instat, and much of the philosophy of R-
Instat is inspired by Instat. Instat included a special menu for the analysis of climatic data and
R-Instat follows this tradition, as well as including another special menu for the analysis of
public procurement data.
You can find out more about the ADI (R-Instat) Team at:

http://r-instat.org/

2
R-INSTAT INTRODUCTORY TUTORIAL PART 1

2. EXPLORING R-INSTAT

This section provides an initial set of examples to help you become familiar with R-
Instat and its general features.

STARTING

Once installed and opened you should see the screen like this:

Fig. 1: R-Instat main Interface

A FIRST TASK – IMPORTING DATA FROM THE LIBRARY

⮚ Go to File > Open From Library or click on the Open from library… shortcut in Fig. 1.
⮚ Click on the From Package dropdown and choose ggplot2.
⮚ Choose the first example, diamonds as shown in Fig. 2.

3
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig. 2. Using a library dataset

⮚ A second Help button is now enabled, just below the list of datasets. Click on
that button to get further information about the dataset. (The diamonds
dataset is used by Hadley Wickham, the author of ggplot2, for many examples
in his own documentation.)
⮚ Now return to the dialog, select the diamonds dataset again and press OK.

Fig. 3 The diamonds data

The data, in Fig. 3, appear that on the left-hand side – looks a little like a spreadsheet.

4
R-INSTAT INTRODUCTORY TUTORIAL PART 1

⮚ Scroll to the bottom of the data to see that it is currently showing the first 1000
of 53,940 rows.

⮚ Use the single arrow to move to the next thousand rows.

⮚ Or the double arrow to go straight to the end of the data set, Fig. 4.

Fig. 4. Viewing a data set

There are 10 columns (variables) of data in this file, of which 7 are numeric and 3 are
categorical. R calls categorical columns factors, and they are denoted by an "F" after the
column name. These categorical columns are ordered, for example the second column,
namely the cut of the diamond’s ranges from Fair to Ideal. Ordered categorical columns are
denoted by "(O.F)" after the column name in R-Instat, Fig. 5.

Fig. 5. Variables

Next, we usually use the prepare menu to organize the data. But here our data is already well
prepared, so we go straight to the describe menu to start the analysis.

5
R-INSTAT INTRODUCTORY TUTORIAL PART 1

SOME GRAPHS

⮚ Go to Describe > One Variable > Graph, Fig. 6.

⮚ Click on the down arrow of the Add button and select Add All. (Or just
select all the columns and then click Add)

Fig 6: One variable graphs dialog

In the dialog in Fig. 6 the radio button changed from Facets to Combine Graph That
is because the selected variables are of different data types; some are categorical
(factors) while others are numeric.

⮚ Press OK to give the results also shown in Fig. 7.

⮚ Click on in the output window to get a bigger view.

6
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig. 7: One Variable Graphs

The analysis in R has detected which variables are categorical and shown them as bar charts,
compared with boxplots for the numeric variables.

Often, the results from using a dialogue can be improved, so you use it again. You could use
the same menu options as in Fig. 6, but there is a quicker way.

⮚ Use the dialogue icon on the toolbar, see Fig. 8. A double click takes you
back to the previous dialogue. Or the down arrow brings up a dropdown list
of the last 10 recently used dialogues.

Fig. 8: Use the toolbar to return to a dialog

7
R-INSTAT INTRODUCTORY TUTORIAL PART 1

You see the dialogue has "remembered" the settings just as you left it when you pressed OK.
This is often convenient.

⮚ This time press the Reset button at the bottom of the dialogue, to clear all the
settings.
⮚ Then select the last 6 variables, so starting with depth, to put into the receiver.
As these are all numeric columns the radio buttons on the right stays on facets, so you can
see what they are!

⮚ Now click OK and in the output window.

Fig. 9: The One Variable Graph dialog again With a faceted graph

This shows a faceted graph, Fig. 9. It doesn’t look very nice. Can you see why? By default, the
y-axis is the same for all the graphs. This is often what is wanted for a multiple graph because
you don't then need the axis to be labelled for each variable.

However, it isn't what we need here. The different variables have very different scales, and
we need to reflect this in the graph. There is an easy fix.

⮚ Return to the same dialogue again.

⮚ Click on the Graph Options button.
You now see a sub-dialogue with just 2 tabs, Fig. 10. One tab allows you to change the type of
graph that is shown, so you could use this to get a histogram instead of a boxplot for
example. The other tab changes how the graph is displayed.

8
R-INSTAT INTRODUCTORY TUTORIAL PART 1

⮚ Press on the tab labelled Display and then click on the Free Scale Axis.
⮚ Press on the Return button and then OK again, to give the graph also shown in Fig. 10.

⮚ Click on in the output window to get a bigger view.

Fig 10: The One Variable graph sub dialog The next graph

Now our boxplots are clear, and each y-axis has its own scale.

9
R-INSTAT INTRODUCTORY TUTORIAL PART 1

SOME SUMMARIES

Often analyses involve numerical as well as graphical summaries.

⮚ Go to Describe > One Variable > Summarise.
⮚ This dialog has three tabs, that summarise the data in different ways. We use the
Default view, Fig. 11.
⮚ Select all the variables again (as you did with for the first use of the Graph
dialogue).
⮚ Press OK to give the results also shown in Fig. 11.

Fig 11: The One Variable Summarise dialog

With some results

10
R-INSTAT INTRODUCTORY TUTORIAL PART 1

These results, just like the graph, are different for the numeric and factor variables. For the
numeric variables they provide the mean together with the same 5 summaries (Min, 1st
Qu., Median, 3rd Qu. And Max) used in a boxplot. For a factor, the frequencies at each level
are given, as we saw earlier in the bar charts.

A MORE AMBITIOUS ANALYSIS

⮚ Go to the Describe > Two Variables > Graph dialog.

⮚ First use the 3 variables: depth, table and price. Then click OK to show the graph and
maximise.

Fig 12: Two Variable Graph

The results show the density of each variable on the diagonal. The 3 correlations are on the
upper part of the display, while the 2-variable scatterplots are on the lower part.

⮚ Return to the dialogue to create a second graph. You do this by double clicking on
the View/Edit Last Dialog icon.

Fig 13: View/Edit Last Dialog Icon

⮚ Add the Cut factor to the Colour (Optional) receiver.

11
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig 14: Two variable Graph 2

It looks quite nice in colour! And now the correlations, and the scatterplots are given
separately for each level of the cut factor.

Let’s make one more graph.

➢ Return to the dialog again, either through the menus or using the View/Edit Last
Dialog icon. This dialog allows both factors and numeric variables. We add the factor
variables colour and clarity. What do we get then?

12
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig 15: Colour Graph with Numeric and Factor Variables

Once you have the graph,

➢ Click on the blue maximise link in the Output window Fig. 16.

Fig. 16: Maximise

13
R-INSTAT INTRODUCTORY TUTORIAL PART 1

CORRELATIONS

The correlations were given as a sort of extra in the Two-Variables > Graph dialogue above.
We now look at more correlations and in more detail.

We examine the relationship between the data in the x and z variables. The diamond data R-
Help file, accessed earlier via the Import From Library dialog, explained that:

• x - length in mm (0–10.74)
• z - depth in mm (0–31.8)
➢ Go to the Describe > Two/Three Variables > Correlations (Fig. 17)
➢ Put the variables x and z into the 2 receivers. R-Instat needs to know both
variables before doing a correlation, so the OK button is not enabled until they are
entered.

This dialog has an Options sub-dialogue (Fig. 17) to add further options choices.

➢ Click on it. Select the Pairwise Plot, then click Return.

Back on the main dialog.

⮚ Click Ok and Maximise the graph in the output window.

Fig. 17: Correlations Dialog and Sub-Dialog

The resulting pairwise plot (Fig. 18) shows the correlation between the length and depth of
the diamonds is very high, which is not surprising.

14
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig. 18: Pairwise Plot

But there are some very odd values, these have been circled in red in Fig. 18. It looks as
though some diamonds have a depth of zero, which is obviously impossible and there is at
least one value that is 3 times larger than any other. With 53 thousand values it is perhaps not
surprising that there are some oddities to investigate.

15
R-INSTAT INTRODUCTORY TUTORIAL PART 1

OUTLIERS

For this investigation,

➢ Go to the Describe > One Variable > Frequencies (Fig. 19)

➢ Select the z variable.
➢ Ok is now available, so press it.

Fig. 19: Frequencies Dialog and Results

Examining the results (Fig. 19), shows 20 diamonds have a z, or depth, of zero, and no others
are less than 1mm. This is difficult to understand. At the other end the second largest
diamond is 8mm deep, then one is 31.80. This is very odd. Perhaps it should have been
3.18?

What to do about outliers is an important topic in data analysis. Here we look briefly at the
results without them, to see if excluding them is a good solution.

The ability to filter data is an important and powerful facility in R and hence in R-Instat. It is a
slightly more complicated dialogue.

➢ Right-click and choose Filter Rows...

➢ In the Filter dialog (Fig. 20) choose to Define New Filter (Fig. 21).
➢ In the sub-dialogue choose z and then give the condition as > 0 and add the condition.
➢ Then use z again, put < 10 and add this condition as well.
➢ Then return and accept as a filter by pressing Ok.

16
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig. 20: Filter Dialog Fig. 21: Sub Dialog

The row numbers on the left, in Fig. 22, are now in red to indicate a filter is in operation. At
the bottom the numbers indicate there are now 21 values less – as expected from the results
earlier.

Fig. 22: Filter Applied

➢ Return to the correlation dialogue. It remains as before, which is convenient, so

simply press Ok and then Maximise from the output window.

The results look better, though there remain several more possible outliers to investigate.
The change in the display from omitting so few observations, shows the importance of paying
attention to checking the data quality, before doing too much in an analysis.

17
R-INSTAT INTRODUCTORY TUTORIAL PART 1

Fig. 23: Original Pairwise Plot Fig. 24: Pairwise Plot Some Outliers Omitted

3. REFLECTIONS

It is easy to follow instructions without being clear on the main points being covered.
We list some of those points here:

● File > Open from Library was used to choose a data set for analysis. Similarly,
the File > Open dialogue can be used to import your own data.
● The data were well organised and ready for analysis, so we used the Describe menu.
● Initial exploration of data often starts by examining variables one at a time. So,
we started with the Describe > One Variable > Graph dialogue.
● The first step in almost every dialogue is to select the variables for analysis.
● We often had to return to a dialogue to refine the analysis.
● The dialogues "remembered" their last settings, so small changes were quick to do.
● Some dialogues have sub-dialogues that give more options.
● On the statistical side it was very easy to produce "multiple graphs". They are useful.
● Next, we wonder whether you consider Fig. 14 and 15 to be graphs or tables?
They have characteristics of both, and the merging of these ideas is one
reason the menus in R-Instat are Describe and Model, rather than the more
traditional Graphs and Statistics.
● Now another statistical point. Earlier in this tutorial we claimed that the
diamonds data was well prepared, and it has been used repeatedly by others.

18
R-INSTAT INTRODUCTORY TUTORIAL PART 1

So we were able to omit the Prepare menu and start immediately with the
Describe menu for our analyses.
Fig. 25: Describe Menu Fig. 26: Prepare menu

• Then the correlation analysis, in Fig. 23, indicated that there were some
oddities in the data, that perhaps should have been examined first. We
therefore went back a step to the Describe > One Variable > Frequency
dialogue (Fig. 25) to examine the outliers in more detail. Most of the
dialogues in that menu do “double-duty” and are also available in the Prepare
> Check Data Menu. Just because the data has already been used extensively
by others is not a sufficient reason to omit your own data checking.
● Our main purpose in this tutorial is for users to start learning how to use R via
R-Instat. But we are immediately able to combine this, with some statistical
ideas. We can be productive in teaching and learning statistics at the same
time as we master R-Instat.

19
R-INSTAT INTRODUCTORY TUTORIAL PART 1

4. NEXT STEPS

You can continue exploring the Describe menu with this data set and produce more
tables and graphs that explore the data. The next part of the tutorial introduces
dialogues in the Prepare menu using a second data set from the R-Instat library.

5. FEEDBACK AND REPORTING BUGS

R-Instat is still under active development with improvements and new features
planned for future versions. We appreciate your feedback to help us improve R-Instat.
There are several ways you can provide feedback:

1. For general feedback contact us via email at:

[email protected].
2. Our issues page on our GitHub account can be used to report specific
bugs or suggestions and this is the most direct way to contact the
development team. Our issues page is publicly visible to anyone. It can be
accessed here: https://github.com/africanmathsinitiative/R-Instat/issues.
Click the green New Issue button on the right side to send your message.
When reporting a bug or problem, it’s most helpful if you can be as specific as
possible, and detail how to reproduce the bug. Attach data if possible. You may
be able to include a capture of the completed dialogue. Ideally, (if this is
possible) also paste the R code from the Output window or from the log file.

R-Instat Team, African Data Initiative

20