Week 1: Chapter 3,4,6&7 Flashcards
Correlation
Changes in one variable are associated with changes in the other. Correlation simply measures the degree and direction of the relationship between variables.
Causation
The direct influence of one variable on the other, changes in the cause variable directly cause changes in the effect variable.
Why do we use log transformations?
- Normalize exponential and skewed data
- Allows for a more accurate interpretation of value
- Shows the non-linear relationship in the model
Level-Log regression
A 1% increase in x is associated with a B1/100 unit change in y, CP
Log-Level regression
A 1 unit increase in x is associated with a B1 * 100 percent change in y, CP
log-log regression
A 1% increase in x is associated with a B1 percent change in y, CP
Linear Probability Model (LPM)
Tests the probability of a binary event
MLR assumptions (1-4)
- Linear in its Parameters
- Random Sampling
- No Perfect Collinearity
- Zero Conditional Mean
GLR assumption 5
- Homoskedasticity
Write the formula for SSTj in the Var(Bj) for OLS estimators under GM
∑(xij-x(bar)j)^2
Why is R^2 not the best for estimating a model (its limitations)
As the number of independent variables in the model increases, the R-squared tends to increase as well, even if the additional independent variables do not actually improve the model’s predictive power.
How does the adjusted R squared fix this
The adjusted R-squared addresses this issue by adjusting the R-squared for the number of independent variables in the model. (1-R^2j)
CLM assumption 6
Normality
U ~ N (0, σ^2)
GM vs CLM
GM: OLS has smallest variance among unbiased linear estimators
CLM: OLS has the smallest variance among ALL unbiased estimators
