Chapter 1: Pre-design planning
Part 1. Lecture: Design-based inference
Estimators of animal density or abundance are often grouped into two camps: design-based or model-based. Density estimates obtained using SCR are no different. In this video we introduce you to design-based inference, where extrapolation from the sampled region to the unsampled region is based on the idea the sampled points are representative of the unsampled points. The only way to guarantee this is with some form of random sampling. We also show how sampling in mostly favourable areas and then treating the sample as if it was representative will lead to biased estimates.
Part 2. Lecture: Stratified sampling
This video looks at stratified sampling, which divides the survey region up into sub-regions called strata. If the outcome variable (like animal density) is more similar within strata than between strata, this can lead to more precise design-based density estimates. You can also sample some strata (like where animal density is expected to be higher) more intensively than others, provided there is still some random aspect to the designs in each strata.
Part 3. Lecture: Model-based inference
This video looks at model-based inference. In contrast to design-based inference, where unbiased inferences can be obtained so long as the design is randomized, in model-based inference we aim to explain density through a model involving spatially-varying covariates. A good estimate of density needs a good model, and this requires that the full range of any spatial covariates are sampled, not just high or low values.