Bi Mu Fan 709446 Charles Shenton Friday 3:14-4:15 Spot 3011
Question 1 BEERSHARE EDUC HSIZEAVG SQRDHHSIZE WRKCH5
BEERSHARE 1.000000 -0.190567 0.226910 0.274304 0.389749
EDUC -0.190567 1.000000 -0.131605 -0.117346 -0.058629
HSIZEAVG 0.226910 -0.131605 1.000000 0.993436 0.620427
SQRDHHSIZE 0.274304 -0.117346 0.993436 1.000000 0.658682
WRKCH5 0.389749 -0.058629 0.620427 0.658682 1.000000
To explain how the sales share of beer changes the five regressors: percentage of university graduates (EDUC), average household size (HSIZEAVG), squared average household (SQRDHHSIZE), and percentage of working women with children under five (WRKCH5) are chosen. The fitness of SQRDHHSIZE will be further illustrated in question 2.
i. How the variables change with the new viable:
It is reasonable to …show more content…
Interpretation:
The regression model for beer sales share is: From the second hypothesis test, the null test of f-test can be rejected, which means one of the coefficient is non-zero, thus, the model is succeed. In addition, it shows that EDUC, SQRDHHSIZE, HSIZEAVG are all significate at 5% significate level since the t-test can be rejected. EDUC and BEERSHARE are negative related.
When 1 percentage point of EDUC increase, BEERSHRE will decrease around 0.01 percentage point. The unit is percentage point since EDUC is the percentage of university graduates. Same as WRKCH5, whenWRKCH5 increases by 1 percentage point, BEERSHARE will increase by 0.044 percentage point.
As addressed in question 1, the relationship between the household size and beer consumption is not linear, and the first chart above indicate that the t-statistic of HSIZEAVG is greater than 4, therefore it could be considered as a quadratic model. The relationship between the average household size (X1) and beer 's sales share (Y) in the multivariate model is that:
By differentiating, the coefficient of the quadratic regression