1. As in previous exercises, either create a new R script (perhaps
call it graphical_data_exploration) or continue with your previous R
script in your RStudio Project. Again, make sure you include any
metadata you feel is appropriate (title, description of task, date of
creation etc) and don’t forget to comment out your metadata with a
#
at the beginning of the line.
2. If you haven’t already, download the data file
‘loyn.xlsx’ from the Data link
and save it to the data
directory you created during
Exercise 1. Open this file in Microsoft Excel (or even better use an
open source equivalent - LibreOffice is
a good free alternative) and save it as a tab delimited file type. Name
the file ‘loyn.txt’ and also save it to the data
directory.
3. These data are from a study originally conducted by Loyn (1987)1 and subsequently re-analysed by Quinn and Keough (2002)2 and Zuur et al (2009)3. The aim of the study was to relate bird density in 67 forest patches to a number of different environmental variables and management practices. A summary of the variables is: ABUND: Density of birds, Continuous response; AREA: Size of forest patch, Continuous explanatory; DIST: Distance to nearest patch, Continuous explanatory; LDIST: Distance to nearest larger patch, Continuous explanatory; ALT: Mean altitude of patch, Continuous explanatory; YR.ISOL: Year of isolation of clearance, Continuous explanatory; GRAZE: Index of livestock grazing intensity, 5 level Categorical explanatory 1= low graze, 5 = high graze. Add a description of your variables to the metadata you created previously. Clearly highlight which variable is the response variable and which variables are potential explanatory variables.
4. Import your ‘loyn.txt’ file into R using the
read.table()
function and assign it to an object called
loyn
(checkout Section
3.3.2 if you need a reminder). Use the str()
function
to display the structure of the dataset and the summary()
function to summarise the dataset. Copy and paste the output of
str()
and Summary()
to your R code as a
record. Don’t forget to comment this code with a #
at the
beginning of each line (can you remember the keyboard shortcut?). How
many observations are in this dataset? How many variables does the
dataframe contain? Are there any missing values (coded as
NA
) in any variable? How is the variable GRAZE
coded? (as a number or a string?). If you think this will cause a
problem (hint: it will!), create a new variable called
FGRAZE
in the dataframe with
GRAZE
recoded as a factor.
5. Use the function table()
(or xtabs()
) to
determine how many observations are in each FGRAZE
level.
See section
3.5 of the Introduction to R book to remind yourself how to do
this.
6. Using the tapply()
function what is the mean bird
abundance (ABUND
) for each levels of FGRAZE
?
Can you determine the variance, the minimum and maximum for each
FGRAZE
level? Again see section 3.5
of the Introduction to R book to remind yourself how to do this.
7. Now onto some plotting action. Plot a Cleveland dotchart (Section
4.2.4) of each variable separately to assess whether there are any
outliers (unusually large or small values) in the response variable
(ABUND
) or any of the explanatory variables (see Q3).
Produce a Cleveland dotchart of each variable separately to assess this.
If you feel in the mood, output these plots to an external PDF file in
the output directory you created in Exercise 1.
8. If you do spot any variables with unusual observations you will
need to have a think about what you want to do with them (NOTE: do
not just remove them without justification!). If you’re
unsure, be sure to speak to an instructor to discuss your options during
our synchronous practical sessions. One option is to apply a data
transformation to see if this reduces the magnitude of any outlier. The
best thing to do here is to play around with different transformations
(i.e. log10()
, sqrt()
) to see which one does
what you want it to do. Best practice is to create new variables in your
dataframe to store these transformed variables. After you have applied
these data transformations make sure you re-plot your dotcharts with any
transformed variable to double check what the transformation is doing.
Hint: a log10 transformation might help reduce the magnitude
of the outliers for some of the variables.
9. Next, check if there is any potential collinearity between any of
the explanatory variables. Remember, collinearity is a
strong relationship between your explanatory variables. Plot
these variables using the pairs()
function (Section
4.2.5). You will need to extract your explanatory variables from the
loyn
dataframe (using []
) either before you
use the pairs()
function or whilst using it. Optionally,
include the correlation coefficient between variables in the upper panel
of the pairs plot (see section
4.2.5 of the introduction to R book for details) to help you decide
whether collinearity is an issue.
10. Now that we’ve checked for collinearity let’s assess whether
there are any clear relationships between the response variable
(ABUND
) and individual explanatory variables. Use
appropriate plotting functions (plot()
,
boxplot()
etc) to visualise these relationships. Don’t
forget, if you have applied a data transformation to any of your
variables (Q8) you will need to plot these transformed variables instead
of the original variables. Also, don’t forget, you can split your
plotting device up to allow you to plot multiple graphs (Section 4.4)
or again use a function like pairs()
to create a
multi-panel plot. Output these plots to the output directory as PDFs.
Add some comments in your R code to summarise your findings.
11. One of the main aims of this study was to determine whether
management practices such as grazing intensity (GRAZE
) and
size of the forest (AREA
) affected the abundance of birds
(ABUND
). One hypothesis was that the size of the forest
affected the number of birds, but this was dependent of the intensity of
the grazing regime (in other words, there is an interaction between
AREA
and GRAZE
). Use an appropriate plotting
function to explore these data for such an interaction (perhaps a
coplot()
or xyplot()
in Section
4.2.6 might be helpful?). Again, don’t forget, if you have applied a
data transformation to your AREA
variable you need to use
the transformed variable in this plot not the original AREA
variable. Output this plot as a PDF to your output directory and add
some comments to your R code to describe any patterns you observe.
1 Loyn, R. (1987). Effects of patch area and habitat on bird abundances, species numbers and tree health in fragmented Victoria forests. Nature conservation: the role of remnants of native vegetation. 65-77.
2 Quinn, G. P., and Michael J. Keough. 2002. Experimental design and data analysis for biologists. Cambridge, UK: Cambridge University Press.
3 Zuur, A.F., Ieno, E.N. and Elphick, C.S. (2010), A protocol for data exploration to avoid common statistical problems. Methods in Ecology and Evolution, 1: 3-14. doi:10.1111/j.2041-210X.2009.00001.x
End of Graphical data exploration using R Exercise