Академический Документы
Профессиональный Документы
Культура Документы
Question 1:
-
Question 2:
-
Question 3:
Testing the structural change:
We would like to test for the structural change in price of one single commodity spot price, crude oil.
Download the crude oil spot price (WTI or Brent or Diesel) form Quandl from 1985 to 2010.
You can either have the time series plot from the website or import the data to Stata and then make a
graph.
Analyze the price movement by only looking at the graph. What do you see?
Now we would like to conduct a Chow test to check the structural change in the price of oil, which date
in time would you think should be the breaking point?
Now consider the following regression:
= + +
Conduct the Chow test. We need the data for CPI for US economy (monthly basis). Which date would you
chose as a structural break? Conduct the test according to your choice of the structural break and interpret
your results.
Question 4:
- Open the auto.dta data set.
- Regress car type on mpg.
- Run the probit and logit models and interpret the result.
- Compute the marginal effects at means for these three models
- Next calculate the odds ratios. Interpret them.
- Calculate the predicted probabilities for the models. And compare the result of OLS with that of
probit and logit.
Question 5:
a) Open lifeexp.dta data set. And regress life expectancy on gnp per capita and safewater.
b) Check for the heteroscedasticity in a visual sense. In order to do so you may graph the residuals
against the fitted values.
c) Do the heteroscedasticity diagnostics check (white test) and interpret the result.
d) Now rerun the same regression with robust setting and compare your results with part a.
e) Lets try to solve the issue by FGLS. Run the original regression and obtain the residuals and then
calculate an estimate the variance. (Dont forget to get the log). Now run the regression with the
log and the explanatory variables. Then obtain the fitted values. The next two steps would be
getting the exponential fitted values and run the transformed repression model with
1/ (exponential fitted values).