[Return to tutorial page]
Statistics 200: Lab 11-13 (Friday 15 & 22 November, 6 December)
Now you get a chance to show what you have learned, by using Splus to do
For the next two lab sessions you should work on a project involving Splus.
During the final lab session, on Friday 6 December, you will get to explain to
the class what you have done. You need to get started quickly.
Some possible projects
Some of you already have data sets that you want to work with. Try to use the
graphics tools in Splus to show what you have, and what sort of analysis might
- Get hold of some data from the recent election. Try to show patterns
across Connecticut, national patterns, contrasts with polling information, or
whatever. My wife found a table giving presidential votes by town
within Connecticut at
polltrak, but it would take a little bit of work to convert it to a convenient
form. Maybe you could make use of Census data. Maybe you can do
something with the
index might be a good place to start looking.
Write a useful Splus function. For example, it would be nice to have a modified version of
read.table() that didn't suffer from the overhead of writing large temporary
files to extract headers. I think it can be done using scan() more
efficiently. I would like to have a function like:
that would return a data frame. For example, if a file foobar.txt contained
and so on, then I would like
junk <- my.read.table("foobar.txt",sep=":",datatypes="FCFN")
to return a data frame with factor variables junk$court and junk$disq, a
character variable junk$date, and a numeric variable junk$count.
If you want to get fancy you might allow more flexible forms for the datatypes.
For folks in the Linear Models course, we would accept a combined project if it
contained enough material. For example, you might be able to develop some
functions to bootstrap regression residuals and produce some interesting
graphical displays. Talk to Marten about this one.
Find an example where trellis graphics is really a big improvement over the
simpler Splus tools. Create some deviishly clever panel functions.