Sampling techniques for condition reporting

Alison Walster writes through R. A. Reynolds writes

>Can anyone recommend a
>sampling method as it will be impossible to check all the
>objects?

At the Museum of Victoria we  wanted to get some data on the overall
condition of our Science and Technology collection (about 40,000
objects), much of which was unregistered and not condition reported. We
also were unable to assess every object, so we hired a consultant
statistician to develop a genuinely random sampling methodology for
selecting objects to assess, and to interpret the results. The person we
used was

    Malcolm Clarke
    Dept. of Mathematics
    Monash University
    Clayton, Victoria 3168
    Australia

in case you would like to contact him about the work. Reading from
his report the sampling method used was "single-stage cluster
sampling".

I think it is important to get professional advice in developing
random sampling strategies, and in interpreting the results.
Conservators are not trained in the use of advanced statistics, and
if you are using the results of the survey to argue for extra
funding or resources it is useful to know that  your methods are
sound. The survey we did made me aware of a lot of subtleties such as
error margins and confidence limits that really affect the data. For
instance if 80% of objects sampled in one area are in need of
conservation,but only 70% in another area, this appears to be a
genuine difference in the state of those two areas of the
collection. However if the margin of error in you data is 10% in
both cases, this apparent difference could be merely an artefact of
the particular objects you (randomly) happened to sample.

Alison Wain
Museum of Victoria

                                  ***
                  Conservation DistList Instance 10:87
                  Distributed: Tuesday, April 8, 1997
                       Message Id: cdl-10-87-003
                                  ***