Regarding MidTerm Exam ( EPIB681 Data Analysis II )
Updated: 2.50pm, February 22, 2004
ITEM

Comments

exam_samples_2004.pdf 
Sample Questions / guidance on types of questions to
be asked 
logits_to_from_proportions
stopping at odds along the way
( a very small pdf file) 
I prepared this in Excel, as a way to save having to
keep going from logits to proportions, and from proportions to logits, on a calculator.
It may save some time with these types of transformations in the exam. So have a
look at it and see if it could replace your calculator for many such calculations
in the exam  you will still need a calculator for basic addition/subtraction and
multiplication/division.
The accuracy of these "translation tables" between scales is good enough
for exam purposes , where if you need to, interpolation is fine. (and in any case,
the focus will not be on the second decimal place, but on being in the right ballpark/scale).
Indeed they have enough decimal places for what should be used in publications!
The table also shows the relationships between odds and proportion when the proportion
is small, and may help to understand the scales better.
It can also be used to go from x to exp[x] or from x to ln[x], i.e. even when the
purpose is not to convert a beta_hat (a slope on the logit scale) to an odds ratio!
"It used to be" (JH grew up before the electronic age) that these types
of log tables were a standard computing tool.
I will be interested in your reactions to them, and welcome any improvements or corrections. 
leprosy_BCG (pdf) 
This example, taken from Clayton/Hills text, will not
be on the exam.
Notation is different from ours, BUT a good example of a dataset, and an analyis,
for which you might be given a printout, and asked a number of questions.
As in the 4 examples in assignment
7, the key here is interpretation of each parameter,
being able to go backwards and forwards between scales. etc. And the analysis raises
the question of whether age could be modelled some other way. 
more to come.. 
... 
