6.10 Question bank
No answers are given for this section, by design.
Multiple choice
QB 6.1 In the model \(y = \beta_0 + \beta_1 x + u\), the term \(u\) represents
- measurement error in \(y\)
- all factors other than \(x\) that affect \(y\)
- the residual from the fitted line
- the difference between \(\hat{\beta}_1\) and \(\beta_1\)
QB 6.2 The population regression function is
- the line fitted by ordinary least squares
- the conditional mean \(E(Y \mid X = x)\)
- the average of \(y\) in the sample
- the line that minimises squared residuals
QB 6.3 \(E(u \mid x) = 0\) says that
- the residuals sum to zero
- the unobserved factors average zero at every value of \(x\)
- \(x\) and \(y\) are uncorrelated
- the model is linear
QB 6.4 Which of the following are unobservable? (select all that apply)
- \(u_i\)
- \(\hat{u}_i\)
- \(\beta_1\)
- \(\hat{\beta}_1\)
QB 6.5 Minimising \(\sum \hat{u}_i\) rather than \(\sum \hat{u}_i^2\) fails because
- the sum is always negative
- positive and negative residuals cancel
- it cannot be differentiated
- it requires a larger sample
QB 6.6 Treating the conditional expectation function as linear is
- implied by the definition of the PRF
- a modelling choice
- required for \(E(u) = 0\)
- a consequence of random sampling
Short reasoning
QB 6.7 Explain the difference between the population regression function and the sample regression function, and say which one changes when a new sample is drawn.
QB 6.8 A colleague says the error term captures “measurement error in the data”. Explain what is wrong with that, and give three examples of what \(u\) actually contains in a regression of wages on years of schooling.
QB 6.9 Why is \(E(u) = 0\) described as costing nothing, while \(E(u \mid x) = 0\) is the assumption on which everything rests?
QB 6.10 State the zero conditional mean assumption in words for a regression of crop yield on fertiliser use, and give one concrete reason it might fail.
QB 6.11 A researcher regresses district crop yield on rainfall and finds a strong positive slope. Give two reasons the slope may not measure the effect of rainfall on yield.
QB 6.12 Explain why the least squares criterion penalises one residual of 8 more heavily than eight residuals of 1.
Numerical
QB 6.13 A population’s conditional means are \(E(Y \mid X = x) = 12 - 0.5x\).
- State \(\beta_0\) and \(\beta_1\).
- What is the average \(y\) among individuals with \(x = 10\)?
- An individual with \(x = 10\) has \(y = 9\). Find \(u_i\).
- Interpret \(\beta_1\) in words.
QB 6.14 Four individuals with \(x = 6\) have \(y\) values 14, 18, 11 and 21.
- Estimate \(E(Y \mid X = 6)\).
- Compute the four implied values of \(u_i\).
- Verify that they sum to zero, and explain why that had to happen.
- Explain why this does not demonstrate that \(E(u \mid x) = 0\) in the population.