14_310x - Module 01 - Introduction to the Course

May 13, 2021 · View on GitHub

Lecture 0101 - Welcome to the Course

degree of a node -> number of connections a node has
eigenvector centrality -> importance of a node related to how important are its direct connections

Benford's law: the distribution ofthe first digit in a numeric dataset is not uniform, but geometric: the digit 1 is far more common than the digit 9. Using this it is possible to test if datas has been fabricated at art or they are real observations.

If you're very convinced about a particular causal story, you can look at the data in the way that is going to reinforce your belief in this very strong causal story.
So when one looks at data, one needs to use priors and theory to know what to look at.
But one also needs to let the data speak.
And in some sense, tie your hands to look at the data and not keep the part of the data that's good for you and not the part that's less good for you.

Education at aggregate levels is more effective than education as individual level because of "externalities of education", where educated people can share and propagate the knowledge.
Omitted variables could explain why two series looks going together in a trend.

Course program
- probabilities: how to model the processes that might have generated our data
- statistics: how do we summarize and describe the data and and how do we try to go from to date to the processes that might have generated it
- exploratory data anlysis, econometry, machine learning: uncover the patterns between variables to find how variables might be related to each other.
- various regression tests: to find causality in our data
How to do? In practice!:
- using R
- learning experiment design to obtain data
- learn about data sources
How to represent our data and findings ?:
- plots
- tables
- texts

Detailed course program