https://dacss.github.io/DACSS_601_Summer2023_Sec1/posts/AudreyBertin_Challenge9.html

[giscus[bot]]

https://dacss.github.io/DACSS_601_Summer2023_Sec1/posts/AudreyBertin_Challenge9.html

Creating a function

https://dacss.github.io/DACSS_601_Summer2023_Sec1/posts/AudreyBertin_Challenge9.html

[sp-conway]

Nice job Audrey! Though I think the z score should not use absolute value. I suppose it depends how you're using it. For statistical tests, the absolute value is fine, but as a descriptive measure, I would not use absolute value.

[ambertin]

That makes sense and definitely agree with generally using the pure z score. The use of it with absolute value here is definitely a little strange, but it’s a strategy sometimes used within the field of anomaly detection (from cybersecurity), which is where the baseline compared to new value idea that I used in my function came from!

Basically just trying to see how different the new value is from the past without caring about direction, just difference, so you can find the person/potential bad actor with the most extreme change in behavior.

Probably should have clarified that at the beginning since it’s a little unusual. Oops!

[sp-conway]

Oh interesting! Cool to know.