Abstract:
Environmental problems have become a major focus for modern applied statistics, and there are a number of reasons why model selection issues tend to be of particular importance in analysing environmental data. One reason is that there are often vast amounts of data to be analysed, but it may well be that only a small part of the data are relevant to the questions being asked, which often have to do with issues such as longterm trends or extreme values. Thus, the results obtained may well depend critically on the model selected. A second reason for thinking critically about model selection issues in environmental data is the desire to obtain results that are at least compatible with physical models, such as those constructed by atmospheric scientists. These general issues are illustrated by two areas of research from the environmental statistics group at the (U.S.) National Institute of Statistical Sciences, one concerned with trends in tropospheric ozone, the other with the health effects of atmospheric particles.
tr32.pdf