a) You are investigating the presence of seasonal effects in the U.S refrigerator sales. Using quarterly data for the number refrigerator sales from 1978-Q1 to 1985-Q4 (a total of 32 observations) you estimate the following model, with t-ratios given below each estimated coefficient in parentheses:

= 1222.1 + 245.42, + 347.63, 62.14, (20.37) (2.89) (4.09) (0.73)

where t is the number of U.S refrigerator sales in thousands, 1, is a dummy variable that equals 1 for the first quarter and zero otherwise, 2, is a dummy variable that equals 1 for the second quarter and zero otherwise, 3, is a dummy variable that equals 1 for the third quarter and zero otherwise and 4, is a dummy variable that equals 1 for the fourth quarter.

i) Interpret and discuss the estimated values of the regression coefficients on the dummy variables 2, and 4, . How would you interpret the estimated value of the (common) intercept term in this model? (9 marks)

ii) Using the t-ratios given above, determine whether there is evidence of significant seasonality effects in the U.S refrigerator sales. Clearly explain how you reach your conclusion. If you had access to additional estimation results from the model above, is there any alternative way you could test for seasonality effects instead of using the tratios above? (16 marks)

iii) Using two separate equations, write the estimated regression model that applies in the second quarter and the regression model that applies in the last quarter. (8 marks)

b) Define the problem of measurement error in the context of linear regression models. Explain what the potential consequences of this problem are and briefly outline what potential solutions are available. (17marks)