Causal Inference

RD

Assumptions:

  • Continuity: probability of treatment is constant

Effects:

  • LATE

Procedure:

  • Center variable at 0
  • Regress
    • regression with variable that denotes cutoff
    • Local linear nonparametric regression ()
    • Regress with kernel
  • Fuzzy RD
    • Two-stage least squares regression

AB Testing

Comparing two samples

or

reject null.

Other

Chi-squared test

  • Usage:
    • Test for correlation between categorical variables
    • Test for variance

Binomial

  • Assumptions:
    • iid

Product sense

Step 1: ask questions to understand the problem
Step 2: provide a structure
Step 3: identify the users and customers
Step 4: What are the use cases? Why are they using this product? What are their goals?
Step 5: How well is the current product doing for their use cases? Are there obvious weak spots?
Step 6: What features or changes would improve those weak spots?
Step 7: Wrap things up