P2 Ch 5 c4

i)
And for the log

Thus we say the log inc is more getting close to the normal distribution, the inc seems
like not be normal distribution with higher skewness

ii)

Thus we can say that the data after log is getting closer to normal distribution.
iii) This is not right since the normal distribution data after log might not be still normal
distribution .
Iv) that’s right since all of our assumptions of ols are based on the data are close to normal
distribution. IF our y data is kind like unconditional distribution, to change y into log(y) has
higher possibility for more data on the normal distribution, which is become convenient for
us to analysis.
P3 ch5 6
i) Min of score is 0 max is 100
In this case

Min is 19.53 and max is 98.44

ii)
In this case, we have math and eng score is relative to the colgpa, Thus, these variables
are not independent to each other. Thus, it will cause the skewness of the test. That’s
the reason we cant hold assumption 6 for error u
This will cause the β3 getting smaller, might make β3 insignificant effect in the some
level since some of the effect of the acteng is explained by the colgpa.

iii)

Since in this t-value, we have in 5% level the t is still far from the criteria which is 1.96;
Thus the standard error change doesn’t matter to have the change to this
consequence of acteng is not significant effect.

P5 ch6 p1
i)
 In this case we use t-test, the coefficient is 0.3648, so the t-value =5.548>1.96, thus
we can reject H0 β=0， saying the distance to the incinerator significantly effect the
price of house in 5% level.

For each percent of distance in feet farther to the incinerator, the price of house will
increase about 0.36%. which is significant.

iii)
In this case, we have the t-value only the 1.061 which is not significant effect the price
of house. To this end, we say that’s because in estimation 1 we make area, land. And
bath something important into u, that’s this cause the assumption 5 the not change
of variance of u fail, since we take too much variables into the unknown, leading our
estimate biased. After adding the data that we can get more accurate estimation to
say the distance to incinerator is not that much important.

III)
 After we add the square of log(inst), has the lower standard error that’s because when
we take the square, the standard error would be squared and 0.56*0.56 is about 0.32.
What’s more we kan find that t value stayed the same in absolute value since the
square do not change the significance of effect.

iv)
The square of log(dist) is significant when we add it into the estimate
P6 ch6 c 14

T-value is -3.03 that larger than 1.96 so we can reject H0 that beta is 0. But in the two tail of
the test, we can not reject that for longer.
And for the H0 beta=-1 we have the t-value by using 0.497/0.166=2.99 which we can reject
that in the 5%level t-test, but also the same ,we cant reject it when it is two-tail test.

this is the range and value of log(bs), and for the standard deviation,
Thus it has smaller standard deviation, and it has the more closer to get the normal
distribution.
iii) no, this do not fit the regression much better that’s because we need R^2 as the
benchmark to show that, in estimation 1 we have R^2 be 0.04 but only 0.03 in estimation 2
thus the fewer things we can explained by our regression.

Iv ) it will get smaller since the more variables are added into this system, the regression
getting accurate with these variables.

Now we can say the bs is not significantly different to 0 since the t value only the 1.455 <1.96
in 5% level
v) the coefficient of lstaff is the 1% of staff number increasing will decrease the 1%
average salary, that’s negative makes sense because the more staff but the total
money stay the same, thus we have the fewer average salary for each staff in the
school
vi)
this has no statistical significant that’s because no meaning for square the lunch. And
also, it barely changed the regression.
vii) no, since the lunch in estimation in iv we already showed that does not significantly
affect the final salary benefit, it changes a little.
P9
Ch7 12
i)

This saying more people are above the average.

ii)
since the t-value is only 1.477 that we cant reject H0 , saying the portion of male and female
to above the average looks get the same value, there is not sginificant difference in portion
for male and female to be beautiful in 5% level

iii) H0: for male the not beauty has no effect to wage

In this case we can say the male for not beauty has no effect to the percentage of
wage since the t-value only -1.038 which is far from 1.96.
Thus the effect for not beauty is not really significant effect wage in 5% level

H0 for female the not beauty has no effect to waeg

The t-value for female not be beauty also do not effect the wage significantly in 5%
level. The p-value for male and female are 0.29 and 0.54 seperately
iv)
 There is no significant evidence that the female wage is higher for those who are
beautiful. That’s because the t-value only 1.8 though close and significant in 10% level,
but in 5% level, it still not be an evidence for us to say the beauty has more wage.

V yes it will change since the other variables are much more important to effect the wage
directly for one’s ability or social relationship etc. Thus, with these variables out of unknown
u, the estimation for wage will be less biased to make the looks variable less important.

P11 ch8 12
i) they usually compare to some t-value or to see whether the change of standard error
value makes difference for the regression, and the significance.

ii)
The p -value is 2.2e^-16 the value of F test is 334.6 on 3 and 1688DF which we get the F
test shows this has relationship;
iii)
yes there are significant big difference between ols and wls that’s because none of them the
collection change the sign and they all remained statistically significant, but the value changes
a lot.
Iv
In this case the difference is small only the 1.2519% significant to this model
V
WLS is more precise. If we compare with the robust model of significant level for log(exppp) .it
is statistically significant difference between Ols and WLS, which WLs decrease the significant
level which means it is more precise.