search

ImperialCollegeLondon / crim

UROP summer 2020 on statistical machine learning and criminology
MIT License
1 stars 2 forks source link

Frequentist bandwidth selection exploration across cities for incidents data. #3

Open andrewjholbrook opened 4 years ago

andrewjholbrook commented 4 years ago

Using R function calls such as: density(times), bw.nrd(times), bw.bcv(times), bw.SJ(times), bw.ucv.mod(times) and bw.diggle(X), bw.ppl(X), bw.scott(X), bw.CvL(X), where times is the N-vector of time points and X is an N by 2 matrix of locations.

flaxter commented 4 years ago

Task: given a spatiotemporal point pattern (e.g. the locations of shootings in a city detected by an acoustic gunshot locator system), we use a kernel intensity estimator to estimate the background intensity of the point process using a kernel. The key parameter is the bandwidth (also known as the lengthscale) of the kernel. There are a variety of classical frequentist methods for choosing this bandwidth. We are not sure which is best for our setting. Note that kernel intensity estimation is not the same as kernel density estimation, but we will be considering bandwidth selection methods for kernel density estimation. Citation to come for why.

emma-landry commented 4 years ago

I'd like to work on this project, and I can create a Teams channel for it

Bensong0506 commented 4 years ago

I would also like to work on this project and this will be my first choice.

jialingA commented 4 years ago

I'd like to work on this project (my second choice is Stan code)

DianaS1108 commented 4 years ago

I would be happy to work on this project as well :)

anna-ycx commented 4 years ago

I'd like to do this one too!