Problem Set 4
Due on Monday, Nov. 5 5:00 p.m.
Fall 2018
Please round numbers to 4 decimal places. Write down your answers clearly and show all the important steps you used in your calculations. You do not need to provide your programming code in R or STATA.
1 Problem 1 – Empirical Exercise
Download the data set on Guns (with n=1,173 observations provided by Stock and Watson) on NYU Classes (Folder Data). Import the excel file to your statistical software.
Consider the following variables:
• Yit (vio): violent crime rate (incidents per 100,000 members of the population in state i and year t)
• Xit (avginc): real per capita personal income in state i and year t, in thousands of dollars
Consider only observations in the year 1986 throughout the whole prob- lem set!
1.1 Part 1
Consider the following regression model for the year 1986:
Yit=β0+Uit, t=86 (1) 1. Write down the objective function of the least squares method.
Consider only observations of the year 1986.
1
- Derive the OLS estimator βˆ0 for β0 in regression model (1) as a function of {Yit}.
- Estimate regression model (1) using the OLS regression command from your statistical software. Write down the estimate βˆ0.
- Interpret the estimate βˆ0.
- What is the value of the sample variance of the fitted values, s2ˆ ?
Y
- What is the value of the coefficient of determination of regression R2 in model (1)? Interpret the value of R2 and explain in 1-2 sentences why R2 takes on this specific value.
1.2 Part 2
Consider now the following regression model:
Yit =β0+β1Xit+εit, t=86 withεi|Xit ∼i.i.dN(0,σε2) (2)
- Estimate regression model (2) using the OLS regression command. Write down the estimates βˆ0 and βˆ1.
- Write down the standard errors of the estimators βˆ0 and βˆ1.
- Writedownthegeneral(conditional)distributionoftheestimatorβˆ1|Xit (with distribution parameters).
Note: You do not need to derive the variance formula. You can use the corresponding variance formula from the lecture slides. - Write down the unbiased estimator σˆε2 for the noise variance σε2. (Be careful to distinguish between εˆ and ε !)
- Write down the sum of squared residuals (SSR) of regression model (2).
- Compute σˆε2.
- Compute the 90% confidence interval for β1 (and write down the for- mula that you are using).
2
ii
Consider the one-sided test with H0 : β1 ≥ 60.
- Write down the (most critical, conditional) distribution of the OLS
estimator, βˆ1|Xit, under the null hypothesis.
- Write down the test statistic and its distribution under the null hy-
pothesis.
- Compute the test statistic.
- Compute the p-value [make your approach to the question clear by writ- ing down the corresponding probabilities]. What do you conclude?
1.3 Part 3
Consider now the following regression model:
Yit=β1Xit+εit, t=86 withεi|Xit∼i.i.dN(0,σε2) (3)
1. Estimate regression model (3) using the OLS regression command. Write down the estimate βˆ1.
- Write down the standard errors of the estimators βˆ1.
- Compute Y2 .
i i,86
4. Write down the total sum of squares (TSS). What is the difference
between Y 2 and TSS ? Explain why the total sum of squares are i i,86
larger than in model (1).
Hint: When X has no explanatory power in (3), the baseline model (to which the model fit is compared) becomes Y ̃ |β1 =0 = 0 instead of Y ̄ .
3