Chapter_3_Using_RInstat_effectively.qmd

# Using R-Instat effectively
## Introduction

R-Instat is simply a front end to the R programming language, Wikipedia
(2019). R started in the 1990s and consists of a relatively small core,
that is maintained by the R development core team. There are then also
over 12 thousand packages that extend R's capabilities. About 200 of
these packages are included (behind the scenes) in R-Instat.

The front end in R-Instat is written in Visual Basic.Net. This front end
provides the menus and dialogues that are used to run R-Instat. The
default view of R-Instat is shown in Fig. 3.1a. It has 2 windows, one
showing part of the data and the other is for the results or output.

  ---------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.1a***
  ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.1a.png){width="6.158751093613298in"
  
  height="3.271836176727909in"}

  ---------------------------------------------------------------------------------------------------------------------

R-Instat looks a little like a spreadsheet package, but there are
differences. One, shown in Fig. 3.1, is that the results are in a
separate window, rather than on successive sheets. Also, the data shown
in Fig. 3.1a are stored in an R data frame (behind the scenes) and what
you see is often only a small part of these data.

Current spreadsheets have a limit of about 1million rows. This not a
limit in R, (or therefore in R-Instat) where your machine's memory
imposes a limit that is usually larger. However, the effort of
continually copying all the data to the front end would slow R-Instat
and hence (by default) we just show the first 1000 rows of data and the
first 30 columns[^8].

One way to see all the data in the current data frame[^9] is shown in
Fig. 3.1b. Just ***right-click*** on the tab at the bottom and choose
***View Data Frame,*** Fig. 3.1c.

  ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.1b***                                         ***Fig. 3.1c***
  ------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.1b.png){width="2.1374267279090113in"   ![](figures/Fig3.1c.png){width="3.9064785651793525in"
  height="2.0438856080489938in"}                          height="2.8405096237970255in"}

  ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

The same result could alternatively be done through ***Prepare \> Data
Frame \> View data*** as shown in Fig. 3.1d. This gives a dialogue. From
here, as shown in Fig. 3.1e, you can choose any of the open Data Frames.
Then click Ok to again show the data in R, Fig. 3.1c.

  ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.1d***                                         ***Fig. 3.1e***
  ------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.1d.png){width="2.9987478127734035in"   ![](figures/Fig3.1e.png){width="2.9707206911636046in"
  height="2.4772265966754157in"}                          height="2.9802416885389325in"}

  ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

The menu and dialogue in Fig. 3.1d and 3.1e are all part of the
"front-end" of R-Instat. When you click Ok, R-Instat constructs an R
command and sends it to R. The results are then returned to the
front-end.

The output (results) window shows the R command that has been sent, as
well as the output, if any. ***Right-click*** in the ***Output
Window***, Fig. 3.1f, if you wish to turn this off for future commands.

In the data window ***right-click in the name field*** to provide some
common options, Fig. 3.1g. Alternatively, each of these options is
available from the ***Prepare Menu***.

  ---------------------------------------------------------------------------------------------------------------
  ***Fig. 3.1f***                                         ***Fig. 3.1g***
  ------------------------------------------------------- -------------------------------------------------------
  ![](figures/Fig3.1f.png){width="3.0241666666666664in"   ![](figures/Fig3.1g.png){width="2.9767508748906386in"
  height="2.0632163167104114in"}                          height="2.69in"}

  ---------------------------------------------------------------------------------------------------------------

The data and output windows are the most important in R-Instat. There
are four more windows. The use of the two metadata windows is described
in Section 3.2 and the Log plus Script Windows are described in Section
3.3.

## Column and data frame metadata

The R data frames used through R-Instat contain data together with some
metadata. The name of each variable is part of the metadata as is a
label for each variable. R-Instat has data in a set of
***data-sheets***. An R-Instat ***data sheet*** is an R data frame with
added metadata. The added information includes information on key
columns together with links to other sheets. This helps R-Instat keep
track of multiple data frames that are connected, such as the monthly
summaries calculated from the daily data. A data sheet also keeps
information on objects that have been produced and saved, such as graphs
and models.

  -----------------------------------------------------------------------
  ***Fig. 3.2a The toolbar and View menu***
  -----------------------------------------------------------------------
  ![](figures/Fig3.2a.png){width="5.884278215223097in"
  height="2.321505905511811in"}

  -----------------------------------------------------------------------

Use the ![](Figures/Fig3.1h.png){width="0.32342629046369203in"
height="0.29291447944007in"} icon on the toolbar, (Fig. 3.2a) or ***View
\> Column Metadata*** to see the metadata currently associated with each
open data frame. An example is in the top left in Fig. 3.2b.

  ---------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.2b***
  ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.2b.png){width="6.193180227471566in"
  height="3.7589031058617675in"}

  ---------------------------------------------------------------------------------------------------------------------

As with Excel you can open multiple data frames when using R-Instat. The
column metadata shown in Fig. 3.2b has tabs at the bottom, just like the
data (also shown in Fig. 3.2b), so you can check on the metadata for any
data frame. An R-Instat ***data book*** is the set of open ***data
sheets.***

Using ***View \> Data Frame Metadata***, Fig. 3.2b (top right) opens
another window in which each row shows the metadata on a data sheet. The
information in Fig. 3.2b includes the name of the sheet, an optional
descriptive label, and the number of columns currently in the sheet.

If you use ***File \> Save As \> File Data As,*** Fig. 3.2c, at any
stage, then R-Instat saves the whole data book, i.e. all the data
frames, together with all the associated meta data, into a single file.
This file has the extension RDS and this data book can later be
re-opened in R-Instat[^10]. Good practice is usually to work on a single
topic in each data book, i.e. the different data sheets are usually
interconnected. As with Excel, there is nothing stopping you having all
sorts of unconnected data sheets in the same book, but this usually
complicates your work.

Some tasks are made simpler through the Column Metadata window, Fig.
3.2d. You can double-click in the name field of any column to change the
name. You can add or change the label in the same way, Fig. 3.2. For
numeric columns R chooses the number of significant figures to display.
That is also shown in the metadata and can be changed[^11]. In addition,
as shown in Fig. 3.2c, right clicking on the left-hand side gives the
same popup menu of common tasks as is available from the data view.

  ---------------------------------------------------------------------------------------------------------------
  ***Fig. 3.2c***                                         ***Fig. 3.2d***
  ------------------------------------------------------- -------------------------------------------------------
  ![](figures/Fig3.2c.png){width="2.7479910323709538in"   ![](figures/Fig3.2d.png){width="3.1687871828521437in"
  height="2.1073818897637797in"}                          height="3.7432458442694663in"}

  ---------------------------------------------------------------------------------------------------------------

Each Window button in the toolbar, Fig. 3.2e and the Options in the View
Menu, Fig. 3.2f act like an on-off switch. So now use the curly arrow to
reset to the default of the data and output windows side by side.

  --------------------------------------------------------------------------------------------------------------
  ***Fig. 3.2e***                                         ***Fig. 3.2f***
  ------------------------------------------------------- ------------------------------------------------------
  ![](figures/Fig3.2e.png){width="2.7795439632545933in"   ![](figures/Fig3.2f.png){width="2.180667104111986in"
  height="2.5554516622922137in"}                          height="2.7015277777777778in"}

  --------------------------------------------------------------------------------------------------------------

In this section we have added more of the 6 Windows available in
R-Instat. The opposite is also useful. Once in the default layout,
switching off the Data Window gives just the Output (Results) Window. Or
switch off the Output Window to look at more columns of data. However,
in that case remember to switch on the Output window to see any further
results.

## Graphs

Base R has a comprehensive graphics system, and this is used by many R
packages. The grammar of graphics, Wilkinson (2005) is an influential
book and has led to the exciting ggplot package (gg for grammar of
graphics) and graphics system in R. One challenge we had in constructing
R-Instat was to make the ggplot system easy to use. Almost all the
graphs in this guide use this system. One example that does not, is the
adjusted boxplot shown in Fig. 3.5i.

In this section we describe key concepts of the ggplot system and of our
implementation.

One concept is "facets". Fig. 2.4u is an example of a facetted graph,
where there is one facet for each month. The default is for the x and y
scales to be the same for each graph[^12], so the months can be compared
easily. Also, you are not diverted by lots of axis scales for each
month.

Some multiple graphs show different information in each pane, as was
seen in Fig. 2.4m.

Within a graph you can have multiple layers. So, there are 2 layers in
Fig. 2.4r, one for the station and the other for the satellite data. In
Chapter 10 one layer in a map shows the districts in a country and
another shows the position of each climatic station.

In each layer there is a geometric shape, or geom. A geom may be a
point, a line, a boxplot etc.

In R-Instat the common geoms have their own dialogue as shown in Fig.
3.3a. We choose a boxplot in Fig. 3.3b for Tmax at Dodoma by month.

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3a***                                         ***Fig. 3.3b***
  ------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3a.png){width="2.7233081802274715in"   ![](figures/Fig3.3b.png){width="2.960900043744532in"
  height="2.9685115923009624in"}                          height="3.381676509186352in"}

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Every graphics dialogue includes the option to save the resulting graph,
by giving it a name. In Fig. 3.3b we called the graph Tmax_boxplot. This
ggplot graph is now saved as part of the metadata associated with the
Dodoma data frame.

If you don't give a name, then the default name of last_graph is given
automatically. But, of course, that name is overwritten when you do the
next graph.

 -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3c***                                         ***Fig. 3.3d***
  ------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3c.png){width="2.7233081802274715in"   ![](figures/Fig3.3d.png){width="2.960900043744532in"
  height="2.811914916885389in"}                          height="2.83132217847769in"}

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Double-click in the graph, Fig, 3.3d to turn it blue. This also produces
the popup menu and the graph can be copied to the clipboard.

An alternative is shown in Fig. 3.3e. ***Click*** on graph icon
![](Figures/Fig3.1i.png){width="0.315670384951881in"
height="0.25253608923884513in"} in the toolbar to show the last graph in
R's viewer. This only shows a single graph, but the Window can be
resized and, as shown in Fig. 3.3e, there are now many options to save
the graph, or to copy it to the clipboard.

This toolbar option is only for the most recent graph. The ***Describe
\> View Graph*** dialogue, Fig. 3.3f provides the option to view any of
the saved graphs in this way[^13].

 -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3e The R graph viewer***                                         ***Fig. 3.3f Options for reviewing***
  ------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3e.png){width="2.7233081802274715in"   ![](figures/Fig3.3f.png){width="2.960900043744532in"
  height="3.179398512685914in"}                                               height="2.081327646544182in"}

  --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

We use the dialogue in Fig. 3.3f to show a third way to examine graphs;
one that makes use of the excellent plotly package. This is called the
Interactive Viewer in Fig. 3.3f. This opens the graph in a browser
(though you don't need to be online) as shown in Fig. 3.3g.

  -----------------------------------------------------------------------
  ***Fig.3.3g The (plotly) interactive viewer for ggplot graphs***
  -----------------------------------------------------------------------
  ![](figures/Fig3.3g.png){width="6.0783005249343836in"
  height="3.5125656167979002in"}

  -----------------------------------------------------------------------

In Fig. 3.3g the data for October are shown. This demonstrates that the
boxplot shows the median (30.6°C), the quartiles, etc. In addition, you
can hover over any point to find its value and zoom if a part of the
plot is of special interest. This is a system worth exploring, so we add
a second example. This also shows the value of facets in a graph.

Use ***Describe \> Specific \> Scatterplot*** to examine Tmax as the Y
variable against Tmin as the X, Fig. 3.3h.

 -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3h Initial use of scatter plot (geom point)***                                         ***Fig. 3.3i Resulting Graph***
  ------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3h.png){width="2.7233081802274715in"   ![](figures/Fig3.3i.png){width="2.960900043744532in"
  height="2.9541699475065615in"}  
height="2.9195220909886266in"}  

 ---------------------------------------------------------------------------
 
The resulting graph is relatively uninformative, except to demonstrate
there is a lot of data. And you don't need a special graphics system for
a single graph like this. It does show there are a few outlying points
that require closer investigation.

More concerning, as a principle, is that these are time series data. One
property of time series is the seasonality and we could allow for this
by looking separately at each month.

Return to the dialogue in Fig. 3.3h. Click on ***Plot Options*** and add
the ***month*** factor as a facet, Fig. 3.3j. Press Return on the
sub-dialogue. Then make the ***By Variable*** also the ***month***
factor and the L***abel Variable*** the ***Date***. Click OK to produce
the result shown in Fig. 3.3k.

  -------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3j Plot options sub-dialogue***              ***Fig. 3.3kTmax v Tmin by month***
  ------------------------------------------------------ ------------------------------------------------------
  ![](figures/Fig3.3j.png){width="2.646576990376203in"   ![](figures/Fig3.3k.png){width="3.072049431321085in"
  height="2.242176290463692in"}                          height="3.263601268591426in"}

  -------------------------------------------------------------------------------------------------------------

One feature of the graph in Fig. 3.3k is that there is the same x and y
scales for each graph. This is appropriate here and has then the big
advantage that the graph is not cluttered with many axes, so the data
are more easily compared.

As with the overall graph in Fig 3.3i, one aspect to be investigated,
from Fig. 3.3k, is the outliers. This is easily done via the interactive
viewer. So, return to ***Describe \> View Graph*** and choose the last
graph, possibly called scatter_Tmax.Tmin from Fig. 3.3h (or just
last_graph if no name was given).

  -----------------------------------------------------------------------
  ***Fig 3.3l Interactive view***
  -----------------------------------------------------------------------
  ![](figures/Fig3.3l.png){width="6.161750874890639in"
  height="2.648064304461942in"}

  -----------------------------------------------------------------------

The new feature in Fig. 3.3l is that you can now hover over any point
and see the values in more detail. The example shown in Fig. 3.3l is
that on 9 October 1989 both Tmax and Tmin were 15.6°C. Worth checking!
From the graph 15.6°C is very sensible for Tmin but not for Tmax[^14].

Other dialogues also provide useful graphs. As an example, use
***Describe \> Multivariate \> Correlations***, Fig. 3.3m as in Chapter
2.

  ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3m***                                        ***Fig. 3.3n***
  ------------------------------------------------------ ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3m.png){width="2.706126421697288in"   ![](figures/Fig3.3n.png){width="3.306842738407699in"
  height="1.9848436132983378in"}                         height="3.208082895888014in"}

  ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Complete the dialogue as shown in Fig. 3.3n. Click on the Options button
and complete the sub-dialogue as shown in Fig. 3.3o. The resulting
display is in Fig. 3.3p.

For the data this indicates that both Tmax and Tmin have roughly normal
distributions each month. The correlations are quite low. As with the
other graphs more detail can be found from the plotly interactive
display.

 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3o Correlations subdialog***                                        ***Fig. 3.3p Distributions, Scatter plot and Correlations***
  ------------------------------------------------------ ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3o.png){width="2.6294061679790026in"   ![](figures/Fig3.3p.png){width="3.326188757655293in"
  height="2.8274879702537183in"}                                              height="3.118300524934383in"}

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

This is an example of a graphical display where the different panes show
different types of display. Indeed, one pane is numeric. We used to
think that a display in a graph was distinct from displaying tables of
results. But Fig. 3.3p is like a 2 by 2 table, one cell of which
contains numbers, while the others contain graphs.

Finish this section with a little "housekeeping". Several objects
(graphs) have been produced and perhaps are now no longer needed. Use
***Prepare \> R Objects \> Delete***, Fig. 3.3q.

  ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.3q Menu to manage R objects***                ***Fig. 3.3r Delete objects no longer needed***
  ------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.3q.png){width="2.2876793525809274in"   ![](figures/Fig3.3r.png){width="3.70708552055993in"
  height="2.877690288713911in"}                           height="2.5941666666666667in"}

  ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

We choose to delete all the objects, Fig. 3.3r.

## The log and script windows

The final two Windows in R-Instat are for the Log Window and the Script
Window.

The R programming language is very powerful, but with a relatively steep
learning curve. R-Instat gives easy access to a subset of R. However, a
click-and-point system always has limitations. We consider here some
options if these limitations are ever a constraint.

The log file keeps a record of all the R-commands that have been issued
during a session of R-Instat. Use the toolbar option
![](figures/Fig3.1h.png){width="0.288507217847769in"
height="0.2764862204724409in"}, Fig. 3.2e, or ***View \> Log Window***
to open the log file.

In the Log file, the ***right-click menu*** gives various options, Fig.
3.4a, including saving the log file. That action is the same as using
***File \> Save As \> Save Log As***, shown earlier in Fig. 3.2c.

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.4a The Log Window***                         ***Fig. 3.4b The Filter Dialogue***
  ------------------------------------------------------ ----------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.4a.png){width="3.025767716535433in"   ![](figures/Fig3.4b.png){width="2.9556944444444446in"
  height="1.6994028871391076in"}                         height="2.965126859142607in"}

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

The Log file keeps an exact record of what you have done so far in your
R-Instat work. That is useful. If, later, you ever had to justify how
you produced results, then this is your record of what was done.

If you need to ask for help, then we will usually want to relate
anything extra to the best you have been able to do so far. A skilled R
user can see exactly what was done through the log file.

If R is used directly, then we suggest it be used through RStudio. The
log file can be run in RStudio and should produce the same results.
Hence, an analysis could start in R-Instat and then continue in RStudio
if further results are needed that cannot be done through R-Instat.

Sometimes a dialogue can almost do an analysis, but not quite. If small
changes are needed to a command then they can be done, in R-Instat, with
the Script Window

A guide called "Reading, Tweaking and Using R Commands",
[reference]{.mark} allows users to adapt the commands behind any
dialogue. This situation is illustrated with an example.

We show how to use adjusted boxplots (Hubert & Vandervieren, 2008) for
each month with the rainfall data from Dodoma. This is first illustrated
with the ordinary boxplots that are available (and very useful).
However, for the rainfall we would like to have adjusted boxplots
because of the skewness of the data. They are available in the R
package, called ***robustbase*** [reference]{.mark} that is already used
by R-Instat. But adjusted boxplots are not yet available -- at least
when this guide was first written.

First the data are filtered for just the rain days. Then the boxplots
are produced.

With the Dodoma data, as used in Chapter 2, use the ***right-click***
menu (or ***Prepare \> Data Frame \> Filter***) to give the Filter
dialogue, Fig. 3.4b. Note, in Fig. 3.4b, that the Ok button is not yet
enabled. There is also a Script button on the bottom right of the
dialogue, and this is also disabled. Every dialogue has the same set of
five buttons at the bottom. Hence, every dialogue has a ***Script***
button and here we show how it can be used. In Fig. 3.4b, press on the
***Define New Filter*** button to open the Filter sub-dialogue. In the
sub-dialogue, Fig. 3.4c.

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.4c The filter sub-dialogue***                 ***Fig. 3.4d The filter dialogue completed***
  ------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.4c.png){width="3.1675339020122486in"   ![](figures/Fig3.4d.png){width="2.766781496062992in"
  height="2.7031200787401577in"}                          height="2.775612423447069in"}

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

In the sub-dialogue, choose the ***Rain*** variable and make the
condition as ***Rain \> 0.85***. Then click the ***Add Condition***
button, then press the ***Return*** button. A filter has now been
selected and hence the Ok button is now enabled. We don't need the To
Script button at this stage, but note that it has also been enabled.

Now use the ***Describe \> Specific \> Boxplot*** dialogue and complete
it as shown in Fig. 3.4e.

 -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  ***Fig. 3.4e The boxplot dialogue***                 ***Fig. 3.4f Boxplots with width proportional to number of raindays***
  ------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------
  ![](figures/Fig3.4e.png){width="2.8051170166229222in"   ![](figures/Fig3.4f.png){width="3.125in"
  height="3.203753280839895in"}                                               height="3.125in"}

  -----------------------------------------------------------------------------------------------------------------------------------------------------------------

The results are in Fig. 3.4f. They are useful, and pleasantly colourful,
but the number of outliers shows that the ordinary boxplot is not ideal
for data that are as skew as daily rainfall.

So, return to the boxplot dialogue in Fig. 3.4e and press the ***To
Script*** button. The commands are as shown in Fig. 3.4g

+-----------------------------------------------------------------------+
| ***Fig. 3.4g Script window for the boxplot commands***                |
+=======================================================================+
| 1.  Dodoma \<- data_book\$get_data_frame(data_name=\"Dodoma_merge\")  |
|                                                                       |
| 2.  last_graph \<- ggplot2::ggplot(data=Dodoma_merge,                 |
|     mapping=ggplot2::aes(y=Rain, x=month)) +                          |
|     ggplot2::geom_boxplot(varwidth=TRUE, outlier.colour=\"red\") +    |
|     theme_grey()                                                      |
|                                                                       |
| 3.  data_book\$add_graph(graph_name=\"last_graph\", graph=last_graph, |
|     data_name=\"Dodoma\")                                             |
|                                                                       |
| 4.  data_book\$get_graphs(data_name=\"Dodoma\",                       |
|     graph_name=\"last_graph\")                                        |
|                                                                       |
| 5.  rm(list=c(\"last_graph\", \"Dodoma\"))                            |
+-----------------------------------------------------------------------+

The second line -- that starts last_graph is the only line that needs to
change. The equivalent command, from the robustbase package, is as
follows:

last_graph \<- robustbase::adjbox(Rain \~ month, data =Dodoma_merge,
col=\"red\", varwidth=TRUE).

  -----------------------------------------------------------------------
  ***Fig. 3.4h***
  -----------------------------------------------------------------------
  ![](figures/Fig3.4h.png){width="6.059426946631671in"
  height="2.582607174103237in"}

  -----------------------------------------------------------------------

So paste or type that command into the Script Window, which should then
look as shown in Fig. 3.4h. To run the commands from the Script Window
click on the Run All button at the top of Fig. 3.4h.

If there is an error, then correct the typing into Fig. 3.4h. Now, to be
cautious, right-click, see Fig. 3.4h and this permits you to run the
commands one line at a time. It is highly likely that line 2 is the
problem so first go to line 1 and then run line 1, using Run Current
Line, or pressing \<Ctrl\> + \<Enter\>. Do the same for line 2. If that
works then continue with the rest of the lines and the graph, shown in
Fig. 3.4i should appear.

  -----------------------------------------------------------------------
  ***Fig. 3.4i Skew boxplot***
  -----------------------------------------------------------------------
  ![](figures/Fig3.4i.png){width="3.0629844706911635in"
  height="2.541402012248469in"}

  -----------------------------------------------------------------------

This example has used the Script Window for a new command. More often
the script window is for more modest changes, where an option is added
or changed for an existing command.

If the command above were in a new package, that was not currently
loaded into R-Instat, then the code in line 2 would not be recognised.
In that case you would have add an extra line at the top of the script
window to load the package, e.g. ***install.packages(\"robustbase\").***
This assumes you are connected to the internet. It also only need be
done once. This package is then installed in R and can be used on
subsequent occasions in R-Instat without that line being added.

## Don't let the computer laugh at you!

R-Instat is designed to facilitate the analysis of climatic data. This
may be through using the general facilities (File, Prepare, Describe and
Model menus), or through using the special climatic menu that is
introduced in Chapter 4.

R-Instat is simply a click and point front-end to the R programming
language. It is particularly for users who do not wish to spend time
mastering the R language. There is also a special guide for those who
would wish to start with R-Instat and then consider migrating to using R
"properly".

All software has limitations and it is important to recognise when using
R-Instat, or perhaps your favourite spreadsheet package is not the
correct solution for your work. We give an example below. Otherwise you
may fall into the "copy-paste" trap. This is where you do a very routine
job repeatedly, e.g. copy \> paste, copy \> paste, .... Humans are not
good at boring and repetitive jobs and they make mistakes.

Computers, on the other hand, can be programmed to handle repetitive
tasks brilliantly and very quickly. That's why, if the computer watches
you doing copy \>paste, copy \> paste etc, while it has little to do --
then it is probably laughing at you!

Here is an example. It is of a type we discuss in more detail in Chapter
4. Here it is mainly designed to help you to recognise when your
software, and perhaps your skill set is insufficient for the task in
hand.

Climatic data from Garoua, Cameroon have been provided by the Cameroon
Met Service and are now available in the R-Instat library, Fig. 3.5a.

  -------------------------------------------------------------------------------------------------------------
  ***Fig. 3.5a Garoua data in R-Instat***                ***Fig. 3.5b Maximum temperature data***
  ------------------------------------------------------ ------------------------------------------------------
  ![](figures/Fig3.5a.png){width="2.741365923009624in"   ![](figures/Fig3.5b.png){width="3.353319116360455in"
  height="2.056022528433946in"}                          height="2.047866360454943in"}

  -------------------------------------------------------------------------------------------------------------

Fig. 3.5a shows the "standard" layout of the data for R-Instat climatic
analyses. Spreadsheet packages like Excel, recognise this as a
"list"[^15]. Each column in Fig. 3.5a is of a single "type" -- most are
numeric. The different elements each have their own column. In Fig. 3.5a
each row has the data for a single day. In this file there 21915 rows
(days) of data.

The starting point for these data looked very different. Some of the
initial data for Tmax are shown in Fig. 3.5b. In Fig. 3.5b each day of
the month has its own column, so there are 31 columns in the sheet. We
find this to be a common "shape". One additional problem in Fig. 3.5b is
that each sheet has only about 3 years of data, so they are split across
16 different sheets.

The rainfall is in a different "shape" to the temperatures as shown in
Fig. 3.5c and Fig. 3.5d.

  ---------------------------------------------------------------------------------------------------------------
  ***Fig. 3.5c Rainfall data for Garoua***                ***Fig. 3.5d More of the year of rainfall data***
  ------------------------------------------------------- -------------------------------------------------------
  ![](figures/Fig3.5c.png){width="3.0532458442694663in"   ![](figures/Fig3.5d.png){width="3.1601771653543307in"
  height="1.6980621172353456in"}                          height="1.7202580927384077in"}

  ---------------------------------------------------------------------------------------------------------------

For the rainfall, each year is in a separate sheet. At the bottom of
each sheet there are some monthly and annual totals. Hence the rainfall
is a mixture of data and summary values.

There are three obvious ways you can proceed to change the shape of the
data from Fig. 3.5b, c and d into the shape shown in Fig. 3.5a. The
first is to use Excel (or another spreadsheet), the second is to use
R-Instat and the third is to write a program, using R commands -- or
another language, such as Python.

What might you do with a spreadsheet? Here is a possible way:

1)  It would be good to have all the years of data in a single sheet.
    Start with the first year, which is 1999 for the rainfall. Copy just
    the data -- not the monthly summaries -- into a new sheet.

2)  Now go to the next year, i.e. 2000 and copy the data below those of
    1999.

3)  Now go to 2001 and copy and paste again. The computer is starting to
    laugh, and you have a long way to go.

4)  You persevere and have all the years in your new sheet.

5)  Now you want to paste the February data below January, etc. This is
    11 more goes at copy \> paste.

6)  You are bored, so you look briefly at the temperature data in Fig.
    3.5d. You realise, in horror, that you will have 31 copy \> paste to
    do there, with one column currently for each month.

7)  You give up, realising there must be a better way. That's partly
    because you realise the computer is laughing hysterically!

We took the third option and wrote a program in R. In Chapter 4 we
examine whether using R-Instat would be possible (without the computer
laughing at you too much).

However, there is a general message. R-Instat is merely executing R
commands through a click-and-point environment. This approach will
always be limited, for some tasks, compared to using the command
language directly. So, should you find that an analysis requires a lot
of repetition from you, then check whether it is time to use R directly.

When you start using R-Instat it does not mean you need to abandon using
a spreadsheet. Similarly using R directly does not necessarily mean
abandoning R-Instat. When you use R-Instat it automatically generates a
log file with the R commands. This file runs in RStudio. So, you could
then continue using R directly (through RStudio) for those tasks where
R-Instat has been found to be limiting.