ECON 345
Problem set 7
Aarav Aggarwal
V00990817
i) The coefficient on male is 87.75 which means a man
approximately sleeps 1 and a half hours more per week than a
woman, and this is significant at the1% level (t-
male=87.75/34.33=2.56, t critical (2 sided)=2.58) therefore
there is statistically significant evidence that men sleep more
than women all else equal.
ii) One more hour of work (60 mins) leads to 0.163(60)≈ 9.8 less
minutes of sleep. The t stat on totowrk is 0.163/0.018=-9.06
which is highly significant
iii) We need to run a regression without age and age square, as
including them both in the regression age has no effect on the
model.
i) If ∆ cigs is 10 then ∆ log(bwght)=-0.0052x10=0.052=5.2%
lower birth weight
ii) Holding all else equal a white child is likely to weigh 4.5%
more than a nonwhite child, t-white=3, which is statistically
significant
iii) All else equal, if a mother has one more year of education I
associated with a 0.3% increase in birthweight. T-
motheduc=1 which is not statistically significant.
iv) The two equations use different observations therefore, to
conduct a joint significant test we would need to re-
estimate the first equation with the same amount of
observations n as the second.
v) if If ∆ cigs is 20 then ∆ log(bwght)=-
0.0052x20=0.052=10.4% lower birth weight
`
i) the coefficient on male is 3.82, which means holding all
else constant male students score 3.82 percentage points
more the standard error of male is 0.74 than female
students. 95% confidence interval= 3.82± 1.96(0.74) CI:
(3.82-1.45,3.82+1.45) CI: (2.37,5.27) yes this excludes 0
ii) the variable of male holds little explanatory power when
colgpa is included, however the effect is still statistically
significant.
iii) Centring colgpa reduces correlation between colgpa and
male*(colgpa), this aligns closely with the second mode’s
estimate
i) B1 is positive signialing discrimination against minorities
ii) The regression results are ^ approve=0.708+ 0.201 white( 0.018 )( 0.020 )
2
n=1989 , R 0.049 the coefficient on white means that an application
submitted by a white person is 20.1 percent more likely to be
approved (all else equal), this is a statistically significant
difference.
iii) When we add other variables as controls we get b^1 ≈ 0.129 se=0.020
the coefficient has fallen as we are controlling for other factors
that may explain approval but the race effect is still statistically
significant (t-b1=6.45)
iv) Yes there can be predicted values outside (0,1) range and this is
concerning as it can lead to results that don’t make sense
probabilities beyond 1 or 0