Upper Mississippi River Restoration ProgramLong Term Resource Monitoring |
||
The LTRM has historically collected substantial quantities of data from nonrandom locations. These locations should be viewed as ‘judgment locations,’ in that the sampling locations were selected with particular site characteristics in mind. In LTRM parlance and in historical LTRM documents, nonrandom locations have been termed “fixed sites.” Unfortunately, this label has implicitly confused locations that were repeatedly sampled with locations that were selected using nonrandom mechanisms (as with any location, a repeatedly sampled location may be selected probabilistically or otherwise). For this reason, locations historically termed “fixed” by the LTRM may be more helpfully labeled “nonrandom." Response information from LTRM nonrandom locations should not be used for reporting status and trend estimates for other than the sampled sites themselves. This is because the potential biases associated with selecting a nonrandom set of sampling locations are unknown and, consequently, the representativeness of the sampled locations with respect to unsampled locations in the population is also unknown. Further discussion of this topic is provided by Olsen et al. (1999) and Edwards (1998). The statements above do not preclude modeling data from nonrandom locations (where, by ‘modeling,’ we mean deriving inferences based on assumptions other than those associated with the sampling design; see Statistical Models and LTRM Data). Doing so, however, requires that the authority of any model inferences rests on the reasonableness of the model rather, as is often the case with status and trend estimates, on the sampling design. Models of LTRM data may take many forms and are typically motivated by a different LTRM goal than is status and trend estimation. References |