#sampling #balanced #statistics #sample #variables #spatial #methods

envisim_samplr

Sampling methods for balanced and spatially balanced sampling

2 unstable releases

0.2.0 Sep 24, 2024
0.1.0 Sep 19, 2024

#95 in Geospatial


Used in envisim_estimate

AGPL-3.0-only

175KB
4K SLoC

envisim_samplr

Provides design-based sampling methods, with a focus on spatially balanced and balanced sampling designs.

"everything is related to everything else, but near things are more related than distant things"

— Tobler's first law of geography, Waldo Tobler

Balanced sampling utilizes auxilliary information in order to obtain a sample where the Horvitz-Thompson (HT) estimator of the total of the auxilliary information equals the population total of the auxilliaries. This may be very efficient (yield relatively low variance) if there is a linear relationship between the auxilliaries and the variable of interest.[^1]

Spatially balanced sampling uses auxilliary information in order to obtain a sample that is well-spread in auxilliary space, as well as being balanced. The samples can then be seen as a miniature version of the population. This generally yields low variances for the variable of interest, if there is a general relationship between the auxilliaries and the variables of interest.[^2]

[^1]: Grafström, A., & Tillé, Y. (2013). Doubly balanced spatial sampling with spreading and restitution of auxiliary totals. Environmetrics, 24(2), 120-131. https://doi.org/10.1002/env.2194

[^2]: Grafström, A., & Schelin, L. (2014). How to select representative samples. Scandinavian Journal of Statistics, 41(2), 277-290. https://doi.org/10.1111/sjos.12016

Dependencies

~335KB