(10.1) Quant Methods 2

Question 1

4

What are the components of a regression equation

Answer 1

• intercept
• slope - effect on Y of increasing X by 1 unit
• random error
• line of best fit

Question 1

1

Describe an OLS regression

Answer 1

• Minimises squared distance between actual observations + predicted values

Question 2

2

Describe OLS regression notation (population vs sample)

Answer 2

• Population: uses β (true/known parameter)
• Sample: uses b (estimate)

Question 3

Describe the error term in regression

Answer 3

• What is left in Y after removing the intercept (α)
• Reflects variation in Y not explained by X

Question 4

Describe R²

Answer 4

• Proportion of variation in Y explained by the model
• Don’t want too low (model explains little)
• Too high (possible overfitting)

Question 5

Describe standard errors

Answer 5

• Reflect uncertainty around estimates
• Natural consequence of using sample not population
• Small SE = more precise; large SE = less precise

Question 6

Describe the p-value

Answer 6

Probability of seeing an estimate as extreme as ours, given true parameter is 0

Question 7

2

Describe the t-statistic

Answer 7

• Calculated by dividing coefficient by standard error
• Can be used to determine statistical significance

Question 8

What are the 2 types of substantive interpretation?

Answer 8

• Correlation
• Causation

Question 9

2

Describe correlation interpretation

Answer 9

• Comparison between 2 units
• Allows units to differ in other dimensions

Question 10

Describe causation interpretation

Answer 10

• Change within one unit
• Nothing else should vary

Question 11

Describe the fundamental problem of causal inference

Answer 11

• Need to observe same unit with 2 different values of X at same time (impossible)
• So either: observe same unit at 2 different times, or 2 different units at same time

Question 12

Describe omitted variable bias

Answer 12

• A factor affects both X and Y simultaneously
• Contradicts claim that X causes Y

Question 13

What is a way to deal with omitted variable bias?

Answer 13

Add control variables to regression

Question 14

1

What is the limit of adding controls to deal with OVB

Answer 14

• can never be sure all relevant factors are controlled for