Life Data Analysis Project Guidelines

                        Statistics 533


I ask that the projects be done in stages, so that I can provide
feedback along the way. The due dates are given in the important
dates web page.
 
Milestone 1: Description of the data set, including copies of the
pages from the reference source, if it comes from something
previously published (including on the web). Comment on the question
to be answered in the analysis.

Milestone 2: Preliminary analysis of the data.

Milestone 3: Final project.


1. Find an event-time data set from the reliability, engineering, or
scientific literature (but not from a book on the statistics of
reliability or life data).  Because of the nature of this course, I
would prefer to have students analyze reliability life data
(e.g. data from an accelerated or other life test or data from a
field-tracking study).  If you have special interests in another
area of science or engineering (e.g., medicine, biology,
astrophysics, sociology, etc.), you may use data for such areas as
well.

If you have access to other REAL data which have not been published,
they may be used for the project (Outside of those working in
industry, I would not expect many students to have such a data
source). You may, however, be able to plan and conduct your own
experiment.  Censored data are preferred but not necessary.  Data
with explanatory variables (i.e. regression data) are fine, but not
required (such data do, however, tend to be more interesting, but
more complicated). Choose a set of data where observations might be
assumed to be statistically independent.  If you choose a data set
from a published source, your analysis must extend that of the
original source.

Journal articles provide the some interesting data sets.  If you use
one of the journals that I suggest, I would prefer that data come
from newer issues of journals (i.e., the past two or three years),
because I know about many of the data sets in the older issues.


See the project.data.sources web page for some suggestions.

2. Analyze the data.  This is somewhat open-ended.  At a minimum,
construct a probability plot of your data and attempt to fit some kind
of parametric model or models to your data. Try different
models. You do not (and probably should not) report the results of
all models that you fit. Report the interesting results.

3. The written description should, in almost all cases, be less than
3 or 4 pages in length (this is not a constraint, but an
expectation).  Include all interesting and relevant graphics and
tables. Describe the purpose of the study, how the data were
collected, how the analysis proceeded, the results of the analysis,
and conclusions drawn. State necessary assumptions needed for
conclusions and indicate which can and cannot be checked by looking
at the data.  Try to assess whether or not the study was able to
meet its motivating goals and outline potential for making incorrect
conclusions from the study.

4. The should be done in LaTex or Word and should be supplemented
with graphs or tables (extracted from computer output or done by
hand). These should be integrated neatly into the report.  Do NOT
hand in raw computer output.

5. Students may work alone or in groups of size 2 or 3.

6. The project will be graded on a scale of 1 to 10 and will count
as part of the final grade, as indicated in the grading policy web page.