What you hand in for this exercise should be your own work. It is to be handed in by you to
the Statistics Department O ce by Thursday 22nd March, 14.00. You get some of the marks
by bringing (and potentially presenting) an outline of your work at the workshop session on
2nd March.
Before you hand in your work, complete and sign the slip below, cut it o and attach it rmly
to your work.
Please make sure that your name is recorded on the list of students who have handed in their
your work.
Late work will not normally be accepted.
Your submitted work needs to be printed. Do not use handwriting. The length should be at
most ve (5) pages, with a letter size of at least 11pt, including graphs and pictures. Plain
computer output and listings (for which there is no length limit) can be attached to the back
of the work.
Over length submissions will be penalised, normally by ignoring for marking all material beyond
the maximum length.
Non-submission of in-course assessment may mean that your overall examination mark is
recorded as \non-complete", i.e. you might not obtain a pass for the course.
Any plagiarism will normally result in zero marks for all students involved, and may also
mean that your overall examination mark is recorded as non-complete. Guidelines as to what
constitutes plagiarism may be found in the Departmental Student Handbooks and on Moodle.
The lecturer will keep your submission, which the external examiner may wish to see, so make
sure that you keep a copy of your work. You will receive a provisional grade { grades are
provisional until con rmed by the Statistics Examiners’ Meeting in June 2018.
This assessment will be marked anonymously. Please ensure that your name does not appear
anywhere on the submission apart from this page. Use your student number for the rest of the
submission.
I am aware of the UCL Statistical Science Department’s regulations on plagiarism for assessed
coursework. I have read the guidelines in the student handbook and understand what constitutes
plagiarism.
I hereby a rm that the work I am submitting for this in-course assessment is entirely my own.
Signed:
Please print your name:
Date:
The exercise
Design, carry out, analyse and write a short report on a 2k (k > 2) factorial experiment.
Additional information
Do it yourself!
Yes I expect you to actually DO the experiment. There are lots of possibilities. Kitchens o er a lot
of scope. For example, one might study the success rate of popping corn as a function of pan size,
amount of oil, and heat level. There is no need to do anything particularly fancy, although of course
you can if you wish. I’m happy with other \laboratories" than the kitchen, too. I’d be perfectly
happy with a 23 with either replication or a few centre points to estimate error. The point of the
exercise is to make you think about some of the practical issues involved in experimenting.
Perhaps a few words on what might count as plagiarism might not go amiss. I’d be pretty amazed
if you didn’t discuss amongst yourselves ideas for what kind of example to use. If in consequence
the same basic idea gets used by more than one person I don’t see that as a problem. You might
nd an idea in a book, and that is also ne, though you should reference the book in your report,
and do your own experiment, not take the data from the book. What I would see as unacceptable is
exactly the same experiment from two people. You need to decide the details, e.g., levels of factors
and other implementation details, for yourself, not by copying your friend’s experiment.
Finally, because of misunderstandings in the past when students have analysed data from a book,
I repeat the opening sentence: I expect you to do the experiment, i.e., carry out the runs and
collect your own data. If you do not understand what this means please come and ask.
What to write
You are expected to
1. explain the aim of the experiment and what was done,
2. discuss your choice of factors and factor levels and how the response variable was measured,
3. analyse the data (which should be fully listed) and interpret the results in some detail (I also
expect at least one graph of the data),
4. discuss the design and the model assumptions critically, and indicate what more could be done,
5. write a summary containing the most important results that is understandable to a non-
statistician, which should be self-contained, i.e., assume that the non-statistician only reads
the summary.
You can use this list to structure your report, but you don’t have to. However, you should in any
case end with the summary for a non-statistician as the nal section.
Some comments on marking
Each of the ve items listed above will carry 20% marking weight (but see below). Here are some
aspects that I will take into account when marking.
The experiment should be designed in such a way that it can show something that is informative
and not entirely trivial (I know without experimentation that my cake will be burned if I bake
it at 500 C for eight hours).
Regarding the quality of the experiment, I will check whether what was done was reasonable,
given that your time for doing this is limited and it only counts for 20% of your overall marks
for the course. You should think about possible improvements and mention them even if you
don’t have the time to do them. If there is something obvious that should have been done and
you don’t mention it, this will cost you marks.
As said before, you don’t need to use more than three factors, but again, if there are obviously
relevant in uence factors that could have been used without a lot of e ort, I expect you to at
least mention this. There will be some credit for good and original ideas.
I don’t expect you to use blocking or any of the material after Section 6 in the notes, but it’s
not forbidden either and if you do it right and it is reasonable, it will be rewarded.
Regarding the writing, it is important that you explain properly what was done and why.
You should give interpretations in terms of the subject matter of the experiment, not just
anonymous technical statements such as \there is no evidence against the H0".
Regarding the data analysis I will ignore raw computer output that isn’t explained and inter-
preted, so make sure that you explain and interpret all results that seem relevant in proper
sentences (items 3 and 4 above don’t need to be understandable to a non-statistician, so you
are allowed to be a bit technical there). It is acceptable not to use a computer software for
the calculations and to do them manually. However, you should estimate an error variance and
carry out suitable tests and/or con dence intervals.
Of course, mathematical correctness is important, as is correctness of the interpretations. You
don’t need to write down all your calculations (or computer syntax used) but if results are
wrong, it may win you back marks if I see that you did the right thing but just typed a wrong
number in somewhere, although I’d expect you to nd and, if necessary, correct grossly counter-
intuitive results yourself. It should be clear what assumptions you make and you should say
why you think that they make sense.
Please be aware of the length restrictions given on the rst page. Marks may be
subtracted for pointless and irrelevant statements. Particularly it will not help you to show or
discuss several equivalent analyses (such as computing the same things manually and with a
computer; you may do this for yourself checking your results and it could be a useful exercise,
but I don’t want to see this in your submission).
Pre-submission feedback session
Half of the workshop time on Friday 2nd March, 14.00 - 16.00, Birkbeck B36, will be devoted to
a pre-submission feedback session. You will get 5% (part of the 20% for describing the aim of
the experiment) for submitting in advance (by midnight Thursday 1st March) an outline of length
between one and two (reasonably full) pages of the experiment that you’re planning to carry out or
that you have already carried out. This can be made up of material that you plan to submit later
(see \What to Write", particularly the rst two items). The 5 marks are given for providing the
outline, not for its quality.
These outlines will be discussed in groups (\peer-to-peer feedback") with some input from the
lecturer