Lkjhkl Jknj

Economics 3800

Homework 3

Anders Ludvigsen

Part 1)

a)

b)

c)

d)

Ho: homoscedasticity

Ha: hetroscedasticity

P value is low (0.0023) so we reject Ho and concludes there is significant evidence of hetroscedasticity distribution.

d)

P value is low so the error terms are not normally distributed and there is evidence of hetroscedasticity.

2)

lsalary=log(salary)

lexperience=log(experience)

a)

b)

c)

d)

Ho: homoscedasticity

Ha: hetroscedasticity

P value is low (0.0211) so we reject Ho and concludes there is significant evidence of hetroscedasticity distribution. It is better than the first test but still too low.

e)

P value is high so the error terms are normally distributed and there is no evidence of hetroscedasticity using this test. But the BPG test show that there still is hetroscedasticity.

3)

lsalary=100*log(salary)

a)

b)

c)

d)

Ho: homoscedasticity

Ha: hetroscedasticity

P value is hight (0.5839) so we fail to reject Ho and concludes there is no significant evidence of hetroscedasticity distribution (which is good since we now have homoscedasticity.)

4)

a) The third model seems to be the best model since there are no sign of hetroscedasticity.

b) The reason why I choose model 3 is because of 3 things. 1) The scatter plot of residuals do not have a pattern as the 2 other scatter plots. 2) The normality test shows that the data is normally distributed. 3) the BPG test shows that the third model is homescodastic where the 2 other test are heteroscadastic.

3) as seen in model 3, it is clear that experience has a significant influence on salary. For every year of experience salary goes up by: 5.88 % since salary is logged and therefore % change and not unit.

Part 2

1)

2)

3)