You are on page 1of 6

C4.

i)
lpsoda^ = -1,463 + 0,073prpblck + 0,137lincome + 0,38prppov
(0,294) (0,031) (0,027) (0,133)
n=401 R^2=0,087
5% level H0: B1 = 0 H1: B1 ≠ 0
+ t=(0,073-0)/0,031=2,355
+ t(n-k),alpha/2=t(401-4),(0,025)=t(397,(0,025))=1,966
We have: absolute value of 2,355 > 1,966
-> We can reject the null hypothesis that B1 is equal to 0
-> B1^ statistically different from 0 with 5% level

Option 2: We can use the p-value related to prpblck to test the null hypothesis
P-value = 0,018 < 0,05
-> We can reject the null hypothesis that B1 is equal to 0
-> B1^ statistically different from 0 with 5% level

Option 3: We can see that 95% Conf. interval is from 0,0125 to 0,133 and the hypothesis value
is B1=0 is out of the 95% Conf.interval
->We can reject the null hypothesis that B1 is equal to 0
-> B1^ statistically different from 0 with 5% level

1% level H0: B1 = 0 H1: B1 ≠ 0


+ t=(0,073-0)/0,031=2,355
+ t(n-k),alpha/2=t(401-4),(0,005)=t(397,(0,005))=2,588
We have: absolute value of 2,355 < 2,588
-> We cannot reject the null hypothesis that B1 is equal to 0
-> B1^ is not statistically different from 0 with 1% level

Option 2: From the stata output table, we have the p-value is 0,018 >0,01
-> We cannot reject the null hypothesis that B1 is equal to 0
-> B1^ not statistically different from 0 with 1% level

Option 3: We use the 99% Conf interval(reg lpsoda prpblck hseval, level(99))

ii)

The correlation between lincome and prppov is -0,84, which is high and shows a strong
multicollinearity
-> From the stata output, we have the two-side p-value of lincome is equal to 0 and the two-side
p-value of proppov is 0,004 so both of this is < 1% then we can conclude that each variable is
statistically significant at any level

iii)

lpsoda^ = -0,842 + 0,098prpblck - 0,053lincome + 0,052prppov + 0,121lhseval


(0,292) (0,029) (0,038) (0,134) (0,018)
The coefficient on lhseval means that one percent increase in housing value, with other variables
fixed, will lead to an increase in the predicted soda price is about 0.12%
The t statistic is 6,86 -> p-value for lhseval is very small nearly equal to 0

iv)
*When we add lhseval to the function, it makes lincome and prppov individually insignificant at
any level. (from the stata output in (iii), we have p-value of lincome is equal to 0,159 and p-value
of prppov is 0,699, both more than 10%)

*Test the jointly significant of lincome and prppov ( test lincome = prppov = 0)
Option1:
H0: B2 = B3 = 0 H1: B2 ≠ 0 or B3 ≠ 0

-> From the test output, we have p-value is 0,0304 < 0,05 then we can reject the null hypothesis
at 5% level then we can conclude that lincome and prppov are jointly significant at 5% level
Option 2: (Test lại bằng stata)

F = ((0,1839 - 0,1694)/2)/((1-0,1839)/(401-5))=3,518 > F critical value


-> We reject the null hypothesis

v)
*We can see in function (iii)
+ R^2 is almost double
+ lhseval is individually significant, and log(income) and prppov are jointly significant with high
correlation
It is because corr(lincome,prppov) = 0,84, corr(lincome,lhseval)=0,8 which are very high so in
this function we have multicollinearity, however, R^2 is better -> (iii) is more reliability in
determining whether the racial makeup of a zip code influences local fast-food prices

You might also like