todos before final handin.txt

### masterlist of todos before final handin ###

== data ==
* add city level data and rejoin to get another table with more fine-grained regions
  * note that we may only have millennial data for some years
  * need to make sure we’re matching demographic data to the right year
  * alternatively, we can just take one year as the benchmark and make the assumption that consumption/demographics remain constant over a three year window


== analysis ==
* run panel data regression on our old data in order to parse out fixed-effects and fixed time effects

* run single linear regression on new fine-grained region data
  * figure out ways to control for unknown effects
  * instrumental variable analysis 
  * try introducing non linearity

* run multiple regression on new fine-grained region data
  * could be interesting to see how price affects consumption
  * we already have price data, so this is convenient

* run t- and F-tests to determine significance of dependent variables

* run simple kmeans clustering on new fine-grained data (we won’t have time as a dimension anymore, so clustering makes sense now to see how things clump together in the scatterplot)

* MINNA’S SUGGESTIONS: run tests that might refute our result
  * e.g. run it on boomer population, total population too
  * make sure we address all parts of our hypothesis
  * colour-code data in scatterplots by year (or 3D scatterplot)?


== visualizations, graphs, & tables ==
* make a summary table to be included in the poster (indep var, dep vars/regressors, definition of each one, number of datapoints, mean, median, stdev of each one)

* visualize new fine-grained region data (can reuse same script for previous plots)
  * visualize clusters on scatterplots
  * exploratory bar graphs

* visualize best fit lines after running regressions
  * figure out how to plot vertical residual lines


== interpretations & written analysis ==
* reject or accept our hypothesis? why?
* what were our assumptions?
* confounding variables?
* correlation vs causation?
* comb through ellie’s hypothesis testing and visualization slides to figure out what she wants us to include
  * IMPORTANT (SHOULD DO FIRST)
  * discussion of p-hacking and what we did to prevent that


== resources to check out ==
* this site seems to walk through the main components of a linear regression, hypothesis testing question: https://stattrek.com/regression/slope-test.aspx
* https://cs.brown.edu/courses/csci1951-a/slides/ 
  * lectures 8-11