Idea #1 Justin Wong - #32
Conversation
There was a problem hiding this comment.
It would be interesting to see if you could truly identify whether the crime was a result of the sporting event itself. You could create a heat map of crimes that occur in proximity to Madison Square Garden before and after certain sporting events to visually compare the frequency.
|
I remember talking with you about your idea. I like it a lot. One comment I have is you should probably use tree based models in your prediction of crime since they perform better than linear regression. Unless you care more about inference, then it makes sense to use linear regression. Your setup sort of resembles to a differences in differences setting from economics. The idea is, say there are 10 neighborhoods in the city. Neighborhood 4 has an event happening on Sunday night. You want to look at crime in that neighborhood and compare with Sunday vs day before and after. The cool thing is, other neighborhoods can act as a control group because they don't have a sports event happening there. If neighborhood 7 has a game on Saturday, repeat the same analysis for that. If crime keeps jumping on those event days, despite the days being different, as in, sometimes Saturday, sometimes Sunday, and even different times of they year, your results would have a strong base. You can consider creating a "running variable" as in days before and after an event for each reighborhood and groupby summarize crime rates at the running variable level and report averages. I hope this makes sense. Good luck! |
There was a problem hiding this comment.
This is a cool idea, and I think you could use a regression model to compare certain events and how they simultaneously change with the stock market. You could use an interaction coefficient which would account for changes in two variables at the same time.
|
|
|
3 participants
No description provided.