You are on page 1of 3

3.

11
a. Using the three different conditions of iPods we have
i. iPods are new/unscratched
ii. iPod is used but not scratched
iii. iPod is used and has other problems as well
And to put these three categories into a quantifiable measure,
we use two dummy variables, and use the group iPod is used
but not scratched as a base group that therefore does not need
any dummy variable. We can define (N) to stand for New in this
dummy variable circumstance, where it takes a value of 1 if the
iPod is new, and a value of 0 otherwise. Next, we define (UD) as
Used/Defected to signify the group iPod if used and has other
problems as well. In this case, the dummy variable will take a
value of 1 when the iPod is defected and used, and 0 otherwise.
b. In the equation that is given, I think that the Newi coefficient will
have a positive influence on the price of the iPod because if the
iPod is new than it is noticeably nicer and a consumer would
indeed pay more in order to acquire it. Next, I believe that the
coefficient in front of the Scratchi will be negative. This is
because of the fact that when an iPod is scratched, it is
considered damage, and this leads to consumers desiring to pay
less for the product since it is defected. Finally, I would expect
that the sign in front of the coefficient for the variable Biddersi
will be positive as well because when the number of bidders
among an item increases, there is more demand for said item.
This, as a result, will lead to a rise in the price of the product in
my opinion.
c. In my opinion, the data that is collected within three weeks is not
a valid concern. This is because it is incredibly difficult to get
responses from the entire sample of people in the same day.
d. In comparison with our estimations surrounding the coefficients
of the variables that are involved in determining price, we find
that the estimations we made are correct. This is because New
iPods clearly are $54.99 more expensive than the used and
unblemished ones. Also, the variable of the scratched iPods has a
coefficient that is $20.44 less than the coefficient of the used and
unblemished iPods, which shows that defects do lead to a
significantly lower price. Finally, the variable for bidders was also
positive, although in my opinion it is quite insignificant, and only
raises the price of the iPod by about $0.33 per bidder.
e. In the equation above, the value that is missing is the R2 value.
The R2 value is the coefficient of determination that shows what
percentage of the variance in the price, in this case, we can
explain.
4.4

a. If I had to pick between an unbiased non-minimum variance and


a biased minimum variance I would choose the unbiased nonminimum variance. This is due to the fact that OLD is BLUE if and
only if the estimator is unbiased. Therefore, if it were biased,
then our estimator would not be BLUE.
b. Under no circumstances will the answer in the prior part change
no matter how big the non-minimum variance is or how small the
minimum biased variance is. The reasoning in part (a) clearly
states that the only way for OLS to be BLUE is if and only if the
estimator is unbiased, no matter what the variance is.
c. Unfortunately, there is no systematic way of choosing between
the estimates that have a large variation in their biases and the
estimates that have very little variance.
4.7
a. The (Ci) coefficient in the problem shows the relationship
between the percent of the labor force participation and the
percent of the labor force that is non-white in the ith city. The
equation indicates that for every 1% non-white people in the
labor force, the percent of labor force participation will rise by
0.002. The (Di) coefficient in the equation is a dummy variable,
where if the city is located in the south, then the percent of labor
participation will decrease by .80. If the city is located anywhere
else, then this dummy variable becomes 0 and is irrelevant.
b. In the equation above, there is no perfect collinearity between
the independent variables due to the fact that no independent
variables are related to each other.
c. If the estimates of the regression differ from one decade to
another, this doesnt mean that one or both of the estimates are
biased. On the contrary, it actually simply means that the sample
that was used in the old decade was different than the one that
was used in the new decade.
d. I think that this could be a valid comment to be made about the
equation, simply because it seems to have one flaw in the
direct/inverse relationship between Si and the number of years of
schooling completed. I would expect that as the number of
school years completed went up, the labor force participation
percentage would go up as well. Bu tin contrast, it has a negative
relationship with Si, and therefore, it goes down instead. This
leads me to believe that there was indeed an error in some part
of the equation above.
4.11
a. The coefficient of the variable for the DIVSEP means that when
all of the other terms are held constant, a person who was
divorced or separated will consume 2.85 more alcohol than the
person who was not divorced. The sign on the DIVSEP is positive.
This is logical since divorced or separate individuals will more

likely be more depressed and lonely, which can lead to more


alcohol consumption. Therefore, if there was any relationship, it
would be positive.
Second, the variable UNEMP shows that an unemployed
individual consumes on average 14.20 more alcohol than
someone who is employed. The sign for this is positive as well,
and this is logical because an unemployed person would also be
in a depressed and frustrated state because they cannot fins a
job. Therefore, they would indeed drink more. The relative size of
the coefficients also makes sense when compared to each other.
This is because in reality, it would be much more frustrating and
stressful to not have a job, because then you would be unable to
support yourself and your family. Therefore, it makes sense that
this coefficient would be bigger than the DIVSEP one.
b. The meaning of the estimated slope coefficient ADVICE shows
that if a person receives advice from a physician about how to
reduce drinking, then it will increase alcohol consumption by
11.36. This is indeed contrary to expectations because one would
think that when a person received advice about drinking and its
dangers, they would drink less, or not change their drinking. It is
nonsensical that they would drink more because of this. This is a
violation of the classical assumption that the model must be
correctly specified. Since the sign in front of the ADVICE
coefficient is indeed (+) instead of (-) this means that it was
wrongly specified by possibly doing something such as omitting a
variable.
c. The values of the (B) values are different because any and every
estimator that is indeed BLUE will be an unbiased estimator. And
this means that over different samples, you should receive
somewhat similar, but still different values for all of the (B)
values. The (B)s that are collected from these several samples
are considered the sampling distribution of (B).
d. By estimating the equation with 100 samples, we find that the
value of the coefficient of ADVICE is going to be 8.58 instead of
the prior value of 11.36 when you estimate it with 500
observations. The size of the estimated coefficients in the
equation for 100 different observations is different from the size
of the equation that uses 500 observations. The signs are
common though, which is a good sign, except for the one that is
in front of the EDUC variable. This could be due to the fact that
since they were both estimated using different samples, there
may have been an unequal distribution in one of the samples.
e.

You might also like