Although having to derive statistical and graphical summaries in
relation to small data sets is not an ideal thing to have to do, it
may be that you have all the data that are practically available.
There therefore exists a pragmatic need for statistical and graphical
summaries to be produced for these data for presentation to a
non-statistical audience.
The problem I am concerned with relates to hygienist-determined
occupational exposure assessments. The exposures are usually derived
from readings and are subject to many (unknown) errors in addition to
measurement error. Presently hygienists summarise these data by
presenting a geometric mean (log-normal seems to be universally
assumed) of the exposures (sometimes with 95% confidence interval)
together with a box-whisker plot of the data. The data sets can be
as small as a single data item, but typically have 10 to 30 data
points.
Does anyone know whether there exists particular techniques or
guidelines in relation to presenting summaries of small data sets, or
am I best looking at incidental summaries in the literature relating
to statistical (exact) tests for small data sets, or do I just use
common sense in relation to standard exploratory data analysis
techniques?
I will post a summary of responses to the list. Thanks in advance for
your contributions.
Damien McElvenny
======================================================================
Damien McElvenny
Statistician
Epidemiology and Medical Statistics Unit
Health and Safety Executive
Stanley Precinct
Bootle
Merseyside L20 3QZ
Tel: 0151 951 3352
Fax: 0151 951 4703
E-mail: [log in to unmask]
======================================================================
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|