Show HN: 4d-Modeller – R library to make Bayesian spatiotemporal modeling easy | Hacker News Reader

Show HN: 4d-Modeller – R library to make Bayesian spatiotemporal modeling easy

4dmodeller.github.io

20 pointsmnky9800n6 comments

Hi, I've been working with other colleagues at University of Bristol for the last year to build an R-package that makes bayesian spatio-temporal modeling easy. That is, if you have some problem like predicting sea level rise over time, disease transmissions, etc., this package may be useful for you. It centers on a collection of Shiny apps that guide you through using the various pieces of spatio-temporal modeling packages like R-INLA or inlabru. Hopefully people find this useful. I think the nice part is that we really tried to design the package so you can add it to your modeling tool kit without having to learn a lot about R to be useful with the tool. Hopefully we succeeded.

We will have a hackathon for this model in March 25-27:

https://4dmodeller.github.io/4DM_Hackathon/

At the hackathon we will implement new features or apply the model to new applications.

Open articleView on HN

Tangential question: I would like to compute many different mobility measures from GPS data (beyond buffer zones and convex hulls). I was wondering if anyone can recommend Python/R packages or standalone software that has many mobility measures built-in, is easy to use, and is not ArcGIS?

I think R has ALTREP to support this. It’s on the C API side and your code is pure R, but I there may be online packages which let you create vectors where some of the contents are stored on-disk.

Looks really promising! How do you handle larger-than-ram datasets?

That's a good question and our solution in the past has been to buy larger compute (e.g. https://www.globalmass.eu/). There is no current out of ram solution however it could be an focus of the next Hackathon.

Check out Apache arrow[0], it chunks data into independent categories (e.g. date) and then indexes it contents for quick paging. It works well with the tidyverse, but not perfect, which is why I was curious as to how you guys were doing it.

1: https://arrow.apache.org/docs/r/articles/data_wrangling.html

Cool, I created an issue for it and will look into it post holidays

https://github.com/4DModeller/fdmr/issues/298#issue-20549082...

Reply on news.ycombinator.com