Econometric Analysis Due January 24th, 2018, in class at 1:30pm
Instructions: While cooperating on the assignment is encouraged, pla-
giarism is not. I will only accept hand written assignment submitted in
person. DO NOT SUBMIT YOUR ASSIGNMENT ELECTRONICALLY.
No late assignment will be accepted; late assignments (after 1:30pm on the
due date) will receive a mark of zero. Show your work as no marks will be
allocated for the nal answer alone.
Use Stata or R to do these. If you choose to use another software, please
get my approval at least a week before the assignment is due.
Question 1
Using the International Youth Survey (2006) dataset available on odesi:
(a) Regress, using OLS, the number of times the individual visited a
doctor on the number of group ghts they participated in, gender, whether
they have smoked a whole cigarette and whether they have ever drank beer,
wine, coolers or hard liquor.
(b) Perform. a t-test on the equality of the coe cient for gender to -1
using the output obtained in (a).
(c) Together, how much of the variation in the number of MD visits do
alcohol and tobacco use explain?
(d) If an observation in the sample is such that it is female who has had
2 injuries/accidents in the past year and has never smoked or drank, what
is the predicted number of physician visits?
(e) This observation is such that they visited the MD is 4 times in the
past year. Calculate the residual.
(f) The dependent variable is bounded at 0. Perform. steps (a) and (b)
again using this information. How do your results change?
(g) Using a probit model and the method described in class, determine
whether you are using the right model in step (f).
(h) The dependent variable is a count variable. Should you change the
method you used in (e)? If so, does it change your results and how?
(i) In light of your results in the preceding steps, which model should
you be using and why?
(j) Amongst the models you have used so far in this question, how would
you interpret the coe cient (or do the coe cients represent the marginal
e ect?)? Calculate the marginal e ect for the gender variable and compare
them across models.
(k) Is heteroskedasticity present in these models?