辅导ECN 522、ECN 522留学生辅导、c/c++程序语言讲解、辅导Java/Python 解析C/C++编程|调试Matlab程序
- 首页 >> 其他 ECN 522 Spring 2018
Problem Set 6
Due in class by 12:30pm March 27th.
Problem 1
This problem uses the geocol 24-26.dta on Blackboard, which contains wage and other information about
24-26 year old males from a 1976 survey. You can use the command "describe" to check the deanitions of
the variables.
1. Regress wage on ed, black, and exp variables and interpret all of the regression coe￠ cients.
2. Generate a new variable nonblack=1-black in Stata. (use the command "generate") Run a regression
of wage on ed, nonblack and exp and comment on the results. What happens to the intercept of the
regression? Now run a regression of wage on ed, exp, and both black and nonblack dummy variables
and comment on the results. Next, run the same regression but omitting the intercept. Can you
interpret the coe￠ cients on the black and nonblack dummy variables? the di§erence between them?
Be precise.
3. Variable kww is the personís test score on the "Knowledge of the World of Work" test. Regress wage
on ed, black, exp, and kww. Comment on what happens with the coe￠ cient on ed. Variable kww2 uses
the same test answers as the variable kww but weights test questions slightly di§erently. Calculate the
correlation between kww and kww2. Regress wage on ed, black, exp, and kww2. Then regress wage
on ed, black, exp, kww, and kww2 and comment on the results, in particular on the standard errors
of the coe￠ cients on kww and kww2.
4. In the regression of wage on ed, black, exp, kww, kww2, test the null hypothesis that the test scores
kww and kww2 do not help explaining wage. Do this by using heteroskedasticity robust F-statistic
and the "test" command in Stata, and then by using homoskedasticity only formulas for the F-statistic
using R2 of restricted and unrestricted regressions. Finally, using the "test" command in Stata, test
the hypothesis that the null coe￠ cients on kww and kww2 are equal.
1
Problem Set 6
Due in class by 12:30pm March 27th.
Problem 1
This problem uses the geocol 24-26.dta on Blackboard, which contains wage and other information about
24-26 year old males from a 1976 survey. You can use the command "describe" to check the deanitions of
the variables.
1. Regress wage on ed, black, and exp variables and interpret all of the regression coe￠ cients.
2. Generate a new variable nonblack=1-black in Stata. (use the command "generate") Run a regression
of wage on ed, nonblack and exp and comment on the results. What happens to the intercept of the
regression? Now run a regression of wage on ed, exp, and both black and nonblack dummy variables
and comment on the results. Next, run the same regression but omitting the intercept. Can you
interpret the coe￠ cients on the black and nonblack dummy variables? the di§erence between them?
Be precise.
3. Variable kww is the personís test score on the "Knowledge of the World of Work" test. Regress wage
on ed, black, exp, and kww. Comment on what happens with the coe￠ cient on ed. Variable kww2 uses
the same test answers as the variable kww but weights test questions slightly di§erently. Calculate the
correlation between kww and kww2. Regress wage on ed, black, exp, and kww2. Then regress wage
on ed, black, exp, kww, and kww2 and comment on the results, in particular on the standard errors
of the coe￠ cients on kww and kww2.
4. In the regression of wage on ed, black, exp, kww, kww2, test the null hypothesis that the test scores
kww and kww2 do not help explaining wage. Do this by using heteroskedasticity robust F-statistic
and the "test" command in Stata, and then by using homoskedasticity only formulas for the F-statistic
using R2 of restricted and unrestricted regressions. Finally, using the "test" command in Stata, test
the hypothesis that the null coe￠ cients on kww and kww2 are equal.
1