There are a large number of journals and proceedings that
contain life data that will be suitable for the Stat 533 project.


Your data set should have the following characteristics:

1. Time to failure (or event) should be the (possibly censored) response.
Thus, interval data is suitable.  I would be nice to have some censoring,
but this is not a requirement.

2. There may be one or more explanatory variables (e.g. temperature or
voltage in an accelerated life test).

3. The data set does not have to be large, but try to find one
with at least 5 to 10 failures.


4. Many data sets in the literature are presented in graphical form,
without the actual data.  It is ok to use the graph and to
carefully read numbers from it (perhaps using an enlarged version
of the figure.  In some cases it will be difficult to read data
from the the graphs that are given. At the same time you are reading
off the numbers, you should probably try to write to the author, just
to see if you can get a copy of the real data. This has sometimes
worked for me.


Some things to avoid:

1. Simulated data. They are not interesting.

2. Data from repairable systems (unless there are lots of
systems then, you can analyze the time to first failure only)

3. Results from studies that are given in summary form where it
is impossible to recover the original nature of the 
failure distribution (e.g. tabulated "failure rates").

4. Really old data (e.g. more than 20 years old). Technology is
changing so fast that you would probably be looking
at data from an obsolete product.  Also, data more than 10 years
old is likely to either have been published in one to the
available text books, or be uninteresting.

My experience has been that reliability data sets are sometimes
sensitive, although scientists and engineers are sometimes able to
get permission to publish some of their data.

Reliability data are available from journals in many areas of
engineering.  The most common, however, are in journals
concerned with the reliability of electronics.

From time to time I encounter interesting data sets on the WWW. The
stat 533 home page contains some links that might lead you to
something interesting.

Imaginative experiments are another possibility.

Some journals where good reliability data have been seen
include:

IEEE Transactions on Reliability
   Most papers are on probabilistic reliability and other issues
   that do not use data.  Some articles (regrettably) use simulated data.
   Beware of old data sets; many (especially those in Nelson's
   papers) have subsequently appeared in reliability books.

Microelectronics Reliability
   this one was very disappointing. the quality of the papers was uneven.
   Most papers were are on probabilistic aspects of reliability. 
   The statistical papers that I did find used simulated data.
   There was, however, one very interesting article:

   Dhillon, B.S. and Viswanath, H. C. (1989), ``Bibliography
   of Literature on Failure Data'' {\em Microelectronics Reliability} 30,
   723-750.

   This paper list 367 references which supposedly contain data or information
   on sources of data.

The Proceedings of the Institute of Environmental Sciences
   This is a strange organization, but each proceedings typically
   has some papers on product reliability and some have interesting data.

Electronic Components and Technology Conference Proceedings
   I had good luck here, finding an average of one or two data sets/year
   in the years where I looked.

IEEE Transactions on Components, Hybrids, and Manufacturing Technology
   I have not looked carefully at this one, but did find a couple
   of nice data sets here (I had a reference leading me right to them).

   
Annual Reliability and Maintainability Symposium Proceedings
   There is not a lot of data here, but I have found some in
   adequate graphical form. Beware of all of the repairable system stuff
   here.

Reliability Physics Symposium Proceedings (Annual)
   Mostly low level testing (materials and components). I have seen
   some interesting data sets, most show graphically.

Journal of the American Ceramic Society
   I was also led to a couple of articles in this journal and found
   some interesting data, in graphical form.

In other courses that I teach, students are often successful in
finding data on the web. I would be most interested in receiving
(via email) the URL for any sites that you find that have what looks
like interesting data.