The Beginner’s Path
What is cross-sectional data?
Data collected from different entities at one point in time.
What does OLS do?
It minimizes the sum of squared residuals.
What does R² measure?
The proportion of variance in the dependent variable explained by the model.
If correlation equals zero, have you reached emptiness?
No. The relationship may be nonlinear.
Your p-value is 0.049.Have you achieved enlightenment?
No. You have achieved borderline statistical significance. Remain humble.
What is time series data?
Data collected for one entity over multiple time periods.
What is the difference between correlation and regression?
Correlation measures the strength of a relationship; regression models dependence.
What is multicollinearity?
A strong correlation among independent variables.
You obtained R² = 0.99. This is:
A) Nirvana
B) Overfitting
C) An illusion (maya)
Correct answer: overfitting
But if you confidently say “Nirvana” — +50 for inner confidence
Your coefficient changes sign after adding one variable. What is this?
A) Spiritual awakening
B) Model instability
C) Mercury retrograde
If your answer is A, it is +100
If your answer is B, it is +200
If your answer is C, it is +150
What is panel data?
Data collected for multiple entities over multiple time periods.
When is a logit model used?
When the dependent variable is binary (0/1).
Karmic Risk
-300
You deleted your dataset before the deadline. What is this?
-300
This is karma.
You see *** next to your coefficient. What do they represent?
Statistical significance levels.
Why is panel data more informative than cross-sectional data?
It allows us to control for individual effects and observe dynamics over time.
What is heteroskedasticity?
When the variance of the error term is not constant.
Moment of Enlightenment
+400
What is more dangerous for a researcher?
A) Heteroskedasticity
B) Endogeneity
C) Pride in a high R²
Pride in a high R²
You increase your sample size and suddenly everything becomes significant. Why?
Larger samples reduce standard errors.
What is a fixed effects model?
A model that controls for time-invariant individual characteristics.
What is endogeneity?
When an explanatory variable is correlated with the error term.
Why does correlation not imply causation?
Because of possible omitted variables, reverse causality, or confounding factors.
If your model is insignificant but you deeply believe in it, what should you do?
A) Meditate
B) Recheck the specification
C) Rewrite the data
Meditate
Your regression shows that ice cream sales predict drowning rates. What should you conclude?
A) Ban ice cream
B) Publish immediately
C) Consider omitted variables (like temperature)
C, Consider omitted variables (like temperature)
If you answer A or B with confidence, it is −200 karma points.