<!-- TITLE: Age Period Cohort -->
<!-- SUBTITLE: A quick summary of Age Period Cohort -->
+ [super simple video](https://www.youtube.com/watch?v=t0GuikebSNw)
+ you can't distinguish these. it's literally impossible
+ but we want to
+ **Age effect** affects everyone of that age. People tend to forget about works after 10 years or 20 years, etc, and devalue the knowledge.
+ **Period effect** affects everyone in that year. In 1987 there were less citations total than in 1986.
+ **Cohort effect** affects everyone born in that year. This should be *most* important. Same local intellectual atmosphere.
+ [this talk](https://www.youtube.com/watch?v=j0Cb5g-lx9g) gives a nice introduction
+ [lifelines package in Python](https://lifelines.readthedocs.io/en/latest/index.html)
+ [pyMC3 and categorical variables](http://bebi103.caltech.edu.s3-website-us-east-1.amazonaws.com/2015/tutorials/r7_pymc3.html)
+ [this guy](https://stackoverflow.com/questions/32216901/is-it-possible-to-create-a-hierarchical-model-in-pymc3-using-categorical-random) had a [crazy SO answer](https://gist.github.com/anonymous/c1ada3388a40ae767a8d)
+ [some books](https://www.amazon.com/slp/survival-analysis/b55d5sequ6gm932) -- the social sciences one looks best
+ [MIT lecture notes on survival analysis](https://ocw.mit.edu/courses/health-sciences-and-technology/hst-951j-medical-decision-support-fall-2005/lecture-notes/hst951_sur_anal.pdf)
+ [this notation](https://www.statsmodels.org/stable/index.html) looks nice
The outcome I'm interested in is the probability of death.
I should be able to run this regression. They aren't perfectly colinear in my dataset. I have a huge longitudinal study.
I could just estimate a huge number of fixed effects.
# survival analysis
As I dig deeper, I need to account for people being nested. Maybe survival analysis does better at this?
[first tutorial](https://towardsdatascience.com/survival-analysis-intuition-implementation-in-python-504fde4fcf8e)
it would fix my censored data problem