Andy Wheeler
@crimede-coder.bsky.social
120 followers 59 following 100 posts
Data scientist, with background in criminal justice. Consulting website at https://crimede-coder.com/ Personal blog at https://andrewpwheeler.com/
Posts Media Videos Starter Packs
Pinned
crimede-coder.bsky.social
Giving BlueSky a try, appreciate the follow.

Mostly technical stats and programming content, with a focus on issues in criminal justice.
Chart of weekly motor vehicle thefts in Dallas, with standard errors (Z-scores)
crimede-coder.bsky.social
www.crimrxiv.com/pub/lciht0xi...

And I would have got away with it, too, if it weren’t for you meddling kids": Experimentally falsifying Ghost Criminology, by Ian Adams
Crimversations:
www.crimrxiv.com
crimede-coder.bsky.social
It is soft-launched, but my python package `crimepy` is intended to contain most of the work in my career. github.com/apwheele/cri...
Aoristic analysis of commercial burglaries SNA call-ins and reached graph Time series analysis of outliers using Poisson z-score and seasonal charts. DB Scan hotspots in Dallas
crimede-coder.bsky.social
Book review on the Idea Factory: Bell Labs and the Great Age of American Innovation.

Had no idea Bell had its hand in so many of the building blocks of modern computing. It is also interesting how science + engineering can come together to build great things.

andrewpwheeler.com/2025/09/02/r...
Recommend reading The Idea Factory, Docker python tips
A friend recently recommended The Idea Factory: Bell Labs and the Great Age of American Innovation by Jon Gertner. It is one of the best books I have read in awhile, so also want to recommend to th…
andrewpwheeler.com
crimede-coder.bsky.social
I have updated my Sworn Dashboard, according to the FBI, the percent of female sworn has gone down in 2024 after consistent rises over the last 8 years.

This is mostly due to composition though, more agencies are reporting in the FBI database than in prior years.
crimede-coder.bsky.social
Deep Research tools cite open access work much more often in my experiments than closed source materials. I go over a few examples for Google, ChatGPT, and Perplexity in this blog post.

andrewpwheeler.com/2025/08/28/d...

CJ folks should be posting pre/post prints to plain text HTML (like CrimRXiv)
Paragraph discussing peer review vs other materials cited in Gemini report
crimede-coder.bsky.social
The local newspaper did a writeup of this in Raleigh -- Cisco does a round of layoffs every year (I guess for quarterly reporting to make expenses look lower). There are people who go on the merry go round and like it (they get several months of severance and then get hired back). Free vacation.
criminalerin.bsky.social
Oh that's fine, Cisco isn't important, hardly anything relies on its technology working correctly at all times without fail cuz only a very small percentage of 85% of the whole internet runs on its systems, it's literally fine, stop panicking and go learn to code
Software engineers among hardest hit in latest Cisco layoffs
More than 200 Cisco workers across the San Francisco Bay area will be cut
www.itpro.com
crimede-coder.bsky.social
Pro python tip, I like to develop in the REPL. That way I can use the reload library to edit functions. E.g.

from src import myfunc
test = myfunc(...) # errors out

Then can do

from importlib import reload
reload(myfunc) # edit the source before this
test = myfunc(...) # works
crimede-coder.bsky.social
New blog post, using the SPPT to monitor changes in crime patterns. crimede-coder.com/blogposts/20...

Many of the other techniques (ESRI's emerging hotspots, or KDE maps) tend to be too noisy. This controls for multiple comparisons.
Using the SPPT to Monitor Changes in Crime
Using the Spatial Point Pattern Test to Monitor Changes in Crime
crimede-coder.com
crimede-coder.bsky.social
San Diego sheriff's office is advertising a *work from home* dispatcher position www.linkedin.com/posts/activi...
LinkedIn post on work from home for dispatcher in San Diego
crimede-coder.bsky.social
For folks interested in learning python, grab an epub copy of my Data Science for Crime Analysis with Python, crimede-coder.com/cdcstore/pro....

I made a limited time coupon -- use Ebook20 which lowers the price of the epub to $19.99. Good for the next week.
GIF showing copy-pasting code from Epub
crimede-coder.bsky.social
Using Chicago open data for some example graphs, here is the time series trends for street robberies. Basically as low as it has been going back to 2015
Street robbery time series trends in Chicago.
crimede-coder.bsky.social
Often folks look at large macro patterns of NCVS vs NIBRS. You can use both together to get a more accurate estimate of the total amount of crime in your jurisdiction. This is focused on domestic violence, but can be expanded to any inter-personal crime type.
crimede-coder.bsky.social
The methodology is simple. Imagine you have 100 reported crimes, but only a 50% reporting rate. Your estimate of the total number of crimes (the dark figure) is 100/0.5 = 2. I expand this slightly to estimate different reporting rates for different demographics and over time.
crimede-coder.bsky.social
My peer reviewed paper on up-adjusted NIBRS crime estimates based on domestic violence victimization reporting rates (with Alex Piquero) is now out in Crime & Delinquency.

journals.sagepub.com/doi/10.1177/...
crimede-coder.bsky.social
Schmid also has a book on creating graphics by hand. Incredibly hard work and display of craftmanship.