I'm experimenting with using social media to collect point data of a phenomenon. Obviously, the points are going to cluster in population centers. I'm wondering if any of you know of good methods to try and account for this? The best I can come up with is to join the population of an area (I have fine-scale counts) to the points and then use the inverse of the population as a weight, though I'm not positive it's effective. Any ideas?
I think that approach is quite good in average although can have significativity problems in areas with low population. Perhaps you must aggregate these areas until to obtain enough statistical significance.
This may be related to the small numbers problem, where the variance in a local estimate such as a disease rate depends on the sample size in the area being sampled. More on this, including diagnostic steps, may be found here: http://www.biomedware.com/blog/tag/small-numbers-problem/
Hey guys, thanks for the answers.
Regarding the small-numbers problems... in the past I've used bayesian smoothing, via GeoDa, to reduce the variance. There definitely some block groups with low population (5 people or less) so I think that this is an issue.
- Badges
- Science topic
- Similar topics
- Geoscience
- Geoinformatics
- Geographic Information System