All maps of parameter estimates are misleading

Excerpts

Mapping raw data can lead to spurious spatial features. For example, regions can appear highly variable because of small sample sizes in spatial sub-units (as in the radon example) or small populations (as in the cancer example), and these apparently variable regions contain a disproportionate number of very high (or low) observed parameter values. Mapping posterior means leads to the reverse problems: areas that appear too uniform because of small sample sizes or populations.

— Page 3232

Reference

Andrew Gelman, Phillip N Price “All maps of parameter estimates are misleading” (1999) DOI: 10.1002/(sici)1097-0258(19991215)18:23<3221::aid-sim312>3.0.co;2-m

@Article{gelman1999,
  title = {All maps of parameter estimates are misleading},
  volume = {18},
  issn = {1097-0258},
  url = {http://dx.doi.org/10.1002/(sici)1097-0258(19991215)18:23<3221::aid-sim312>3.0.co;2-m},
  doi = {10.1002/(sici)1097-0258(19991215)18:23<3221::aid-sim312>3.0.co;2-m},
  number = {23},
  journal = {Statistics in Medicine},
  publisher = {Wiley},
  author = {Gelman, Andrew and Price, Phillip N},
  year = {1999},
  month = {dec},
  pages = {3221–3234}
}