2.4 Inference design
2.4.2 Status Inference Design
Overview Status Temporal Spatial Results and next steps
By status, we mean that your indicator is a snapshot during a time interval. The time interval could be seasonal, annual, or cover several years. Some examples include:
- An estimate of the total number of spawners or juveniles
- The average size of size of spawners
- The median time of spawning
- The proportion of spawners that are hatchery fish, or are female
- The proportion of sites that are occupied
- The proportion of time the estimated number of spawners is at or above a benchmark abundance (e.g., as established in a recovery plan)
In developing your goals and objectives (step one), you should have specified the acceptable uncertainty in your indicator estimates (your desired precision) and considered the potential for bias in the selection of your spatial and response designs. Earlier in step 2 you will also have proposed sample sizes and the allocation of sampling effort among sites through your spatial design, and have considered how much uncertainty might be introduced by your response design and the calculation of your metrics.
You should have gathered sufficient information to evaluate whether your spatial and response designs will allow you to achieve your objectives by asking some specific questions and conducting some analyses. At a minimum, you should have information about the variability introduced by your response design, and spatial variability of your metric (unless your spatial design is a single location). You should also have information about the bias that might be introduced by your response design, and in some cases the bias introduced by the selection of your spatial design.
- More Information on Status Inference for Census-based Spatial Designs
-
If your spatial design is a census, the inference design will require defining the procedures used to calculate indicators for the target population based on the metric values from all the sites in the census (or at one site, such as a weir, that is used to enumerate all fish entering or leaving a watershed). Your mock-up should include your best estimates of uncertainty introduced by either your metric’s measurement variability or its bias (i.e., what is the uncertainty introduced by your response design) and the spatial variation in your metric. Does your proposed allocation of sampling effort (sample size) allow you to estimate the indicator with the desired precision. Remind yourself that a census introduces no spatial design uncertainty.
- More Information on Status Inference for Model-based Spatial Designs
-
If you choose a model based design, the quality of your status estimate will depend on the assumptions incorporated into your modeling inferences. Often, geostatistical models (spatial, stochastic models) are used, a simple version of which applies kriging to the data. Typically, several models are available and the selection of which one(s) to use depends on the assumptions required for the model and whether those assumptions are met. Note that you can use model based tools if you use a survey based design or an opportunistic design, however, you cannot use the inference tools applicable to a survey based design for sites selected using a model based approach.
- More Information on Status Inference for Survey-based Spatial Designs
-
If your spatial design is a survey (allowing you to obtain a spatially representative sample of your resource), you must incorporate the additional design uncertainty in your evaluation of your precision. The inference design for estimating a status indicator will also require knowing the properties of the survey design, such as stratification and unequal probability weighting.
- More Information on Status Inference for Opportunistic-based Spatial Designs
-
If you choose an opportunistic design and your objectives are to describe the status of the set of opportunistic sites, then you can evaluate your uncertainty treating the set of sites as a census. Your estimates will apply to the censused sites. However, if you wish to extrapolate your status estimates to a domain (or set of sites) beyond that making up the index set, then you will need to evaluate the bias that might be introduced by your selection of the index sites. Extreme care should be made in making inferences to any areas other than the sites sampled since you do not know how representative your handpicked sites are to anywhere other than the locations sampled. There are few tools available to do this, and the tools available all require you to make some assumptions about how well the sites you’ve chosen represent the target resource.
Review the section on components of variation, as necessary. The evaluation is essentially a mock-up of the methods you would use to estimate your indicators when you have completed your surveys. It is useful to create some example summaries that you might use in your monitoring reports (especially the tables and graphs that you might use for those funding your work and managers who might be using the monitoring results). The mock-up uses your best estimates of the relevant variance components.
Next: Temporal Inference Design | Go Back
