🗝 Key Lessons 

Being able to identify the:

• Limitations of the data
• Limitations of your approach
• Limitations of your tools
• Limitations of the deadline

And being able to communicate those effectively

are ALSO keys to being able to pitch a data-driven story

 correlation != causation

duh...right?

correlation ≠ causation

1. coincidence

3. common cause

2. causality reversed

duh...right?

Communicating Correlation

Correlation doesn't imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing 'look over there'.

XKCD #552

https://fivethirtyeight.com/features/increased-voter-turnout-could-benefit-republicans-or-democrats-in-2020/

Correlation ≠ Causation

What is the role of reporting?

https://fivethirtyeight.com/features/how-much-trouble-is-ted-cruz-really-in/​

Final Project Pitch

• Am I making a causal claim?
  • What is the strongest claim I can responsibly make given my
    • data
    • approach
    • deadline

What is the "best case scenario" headline?

Is this hypothetical "best case scenario" article newsworthy?

• What does the story look like without the causal claim?

Let's discuss some pitches:

Will brighter prospects of congestion pricing spook car buyers and sellers in New York?

Let's discuss some pitches:

The impact of COVID-19 on garbage collection and sanitation in South Bronx

Let's discuss some pitches:

College Football in the Time of Covid: How Home Games Have Impacted College Towns in 2020 and 2021

What is a regression?

(Linear regression)

When do I need a regression?

• When you want observe the relationships between two or more variables...and summary stats / data viz are not good enough tools
   
• When there are a lot of variables that interact with each other
   
• When there are lots of possible things that could explain variance in one variable...
  
  
  ⚠️ Your dataset may not come with the the "inputs" to the regression...this could lead you to make bad assumptions and tell a false narrative! 

Linear Regression

Multiple Linear Regression

Logistic Regression

Etc...other types of models that are not regressions...

What you're trying to measure

vs

How you're measuring it ⚠️

Empirical / quantitative social science on deadline 

- Andrew Flowers (Former Quant Editor @ FiveThirtyEight)

https://www.youtube.com/watch?v=4zLo12JdeOA

Empirical / quantitative social science on deadline 

- Andrew Flowers (Former Quant Editor @ FiveThirtyEight)

https://www.youtube.com/watch?v=4zLo12JdeOA

Data Journalism

• Academic timelines
   
• Build expertise --> publish
   
• Coming up with new findings / methodologies

Academic Social Science

• Working on a news deadline
• Publish --> Expertise --> Publish
• Contextualizing findings, discussing whether they still hold with the latest data
• Using well known and tested methodologies
• Work in consultation with academic social scientists

🗝 Key Lessons

Being able to identify the:

• Limitations of the data
• Limitations of your approach
• Limitations of your tools
• Limitations of the deadline

And being able to communicate those effectively