Academic Integrity: tutoring, explanations, and feedback — we don’t complete graded work or submit on a student’s behalf.

1. A. Quarterly data from 2013 to 2016 produced this estimate of a linear trend

ID: 3172954 • Letter: 1

Question

1. A. Quarterly data from 2013 to 2016 produced this estimate of a linear trend model for gasoline sales (St) across time:

St = 22,902.05 + 117.06t R2 = 0.48

(10.75)

where the number in parenthesis under the estimated coefficient is the t-statistic. Use these regression results to forecast sales for the first quarter of 2017.

B. Using the same set of data, the estimate of an exponential trend model is

lnSt = 10.04 + 0 .0049t R2 = 0.41

(3.15)

Use these regression results to forecast sales for the first quarter of 2017.

C. Which form for the trend fits the data better? (Assess their t-statistic and R2 values.) Also, why might we expect both of them to give poor results?

Explanation / Answer

A. Let the t varies as 0,1,2,3,4 for the quarterky data in the year 2013,2014,2015,2016,2017.

Regression equation is St = 22,902.05 + 117.06t

So, t= 4 for year 2017.

St = 22902.05 + 117.06 * 4 = 23370.29

B. Regression equation is lnSt = 10.04 + 0 .0049t

So, t= 4 for year 2017.

ln(St) = 10.04 + 0.0049 * 4 = 10.0596

St = antilog(10.0596) = 23379.15

C. R-squared of the 1st equation is 0.48

R-squared of the 2nd equation is 0.41

t-statistic of the 1st equation is 10.75

t-statistic of the 2nd equation is 3.15

As, the R-squared and t-statistic of the 1st equation is greater than the R-squared and t-statistic of the 2nd equation, 1st euation (linear trend model) fits the data better.

But both have low R-squared and t-statistic, so both might give poor results. High R-squared shows that the high percent of variation in the “dependent” variable has been explained by the model. The t-statistic is the ratio of the estimate divided by the standard error. So, high t-statistic shows that the estimate is stable and the standard error is low.