Description of this paper

Econ Forecasting Test 3 (2014)


solution. Updated


Question 1.1. If the sum of squared error variation of a variable Y is 28.95 and the sum of squared total variation of the regression is 290.60 the R square value for the regression is: (Points : 3) 11.9 99.5 88.1 78.3 90.0



Question 2.2. What happens to the regression F value if R2 increases holding everything else constant? (Points : 3) The F value decreases. The F value increases. The F value remains the same. It may increase or decrease since it cannot be determined.



Question 3.3. Given the Correlation Matrix below what two major problems can you foresee from the effort to forecast Sales using Price and Advertising as independent variables in multiple regression? Correlation Matrix


Variables Sales Price AdvertisingSales 1.00 .00


Price -.87 1.00 .01 .00


Advertising .45 -.79 1.00 .04 .02 .00


Income .36 .25 .21 .07 .04 .09 (Points : 3) Sales and Income do not have a significant relationship at the 95% confidence level. Price and Advertising may be collinear (Multicollnearity) Price and Sales do not have a significant linear relationship at the 95% confidence level. Both 1 and 2 above. There are not problems indicated with the correlations.



Question 4.4. What statistic is designed to detect multicollinearity in a regression model? (Points : 3) Durbin-Watson Coefficient t-value Standard error Variance Inflation Factor



Question 5.5. What is meant when we refer to multiple regression coefficients as partial or net regression coefficients? (Points : 3) Any independent variable only explains a portion of the variation in the dependent variable. The coefficients have significance levels that achieve the confidence level desired. The coefficients represent the average change in the dependent variable per unit change in the independent variable holding all other things constant. Both a and b above. None of the above.



Question 6.6. What statistic is used to determine if residuals are heteroscedastic? (Points : 3) Durbin Watson


R-square BK squared fit coefficient t-value Variance Inflation Factors 4th Lag LBQ



Question 7.7. The following study estimated demand equation for illicit drugs using a sample size of 4000. The significance level desired for the regression is 5%. The regression output is (Y= Number of days illicit drugs are used):Variable Coefficient Estimate t-ValueIntercept 3.089 17.98Alcohol Prices per liter 0.045 5.93Income ($1000) 0.000057 17.45Male (1 if male) 1.637 29.23Female (1 if female) 1.282 6.35Marital Status (1 if married) -0.807 1.13Age 12 to 20 (1 if between 12 & 20) -1.531 17.97Age 21 to 30 (1 if between 21 & 30) .035 0.51Black (1 if Black) .580 1.84Hispanic (1 if Hispanic) -0.564 -1.03Years in prison for felony drug possession -0.532 -4.72Which of the above variables would you not keep in the regression? (Points : 3) Discard only variables with negative t-values (Hispanic and Years in prison). Delete Marital Status, Age21-30, Black, and Hispanic. Discard Female. 2 and 3 above. Keep all of the variables.



Question 8.8. With time series data when error terms are significantly correlated (or autoregressive) the problem is termed (Points : 3) heteroscedascity. positive serial correlation. multicollinearity. negative serial correlation homoscedasticity



Question 9.9. GE earnings in billions of dollars are estimated using U.S GNP (Gross National Product) in billions of dollars. The regression equation is Y = .065 + .02X. Interpreting the equation (Points : 3) Y =.065, if the value when X is equal to zero. when X increases by one billion dollars the average change in Y is .02 billion dollars. when X increases by one billion dollars the average change in Y is .02 billion dollars plus .065 billion dollars. both 1 and 2 above.



Question 10.10. Which statistic is called the coefficient of determination and why? (Points : 3) t statistic since it indicated which variables determine the variation in Y. R2 since it indicates the share of Y variation determined by X variation. F statistic since it shows the strength of the regression relationship. Mean Squared Regression since it shows the amount on average of Y determined by X.



Question 11.11. When using regression to combine forecast of the same Y variable derived in with various methods how do you determine the best forecasts to include in the model? (Points : 3) Forecast methods that are highly correlated are best to combine with regression Forecasts with low squared residual correlations are best to combine with regression. Forecasts that are most accurate are best to combine with regression. Qualitative forecasts of Y should be excluded from the regression to yield a better combined result. The model that is best has the most significant constant term.



Question 12.12. What approach can be taken in regression to account for qualitative factors or events? (Points : 3) Transformations of X data Stepwise regression Use of dummy variables No approach since qualitative factors cannot be used in quantitative forecasting



Question 13.13. In the analysis of regression residuals what is an indication that heteroscedasticity is present. (Points : 3) High residual values. Megaphone effect in the residual time series or residual order plot. Alternating residuals signs or positive and then negative sign runs in the residual signs. Early lag spikes in the ACFs. Non significant 12th and 24th lag residual LBQ values.



Question 14.14. What is the typical cause of positive serial correlation in business data? (Points : 3) Strong trend that indicates business growth. A lack of an adequate number of data observations. Significant business cycles. Residuals with all positive signs.



Question 15.15. Given the following information below derived from a regression with 200 data observations for each variable are the independent variable coefficients significantly different from zero?Variable Coefficient Std. Error t-values Constant 564 897X1 749 638 X2 .683 .339(Points : 3) X1 and X2 are not significantly different from zero. X2 is significantly different from zero while X1 is not. X1 is significantly different from while X2 is not. Both X variables are significant.



Question 16.16. What statistic can be used to develop a forecast confidence interval around the forecast values? (Points : 3) Standard deviation Standard Error of the Coefficient Standard Error of the Estimate F statistic



Question 17.17. A multiple regression equation is created with 4 independent variables and 24 data observations. The resulting F value is 4.10. If all other statistics for the regression are acceptable would you use the regression to forecast data for your business? (Points : 3) No, because the F values it too low. Yes, because the F value exceeds the table value. Yes, because the F value is positive. It cannot be determined from the above information.



Question 18.18. In determining the best variables to choose for a linear regression model scatter plots can be used. What is an indicator of a good independent variable candidate for regression analysis? (Points : 3) A strong positive or negative relationship between Y and X A linear relationship between Y and X A consistent horizontal relationship between Y and X A consistent vertical relationship between Y and X One and two above.



Question 19.19. Since increasing the number of variables increases R2 why not include every variable in the regression equation? (Points : 3) Yes, you can include as many significant variables as you wish since computation power is the only limitation on regression modeling. No, since the F statistic and Adjusted R2 value degrees of freedom increase and they decreases as a result.


No, since coefficient t-values will decrease as more variables are added. No, since the accuracy of the additional data decreases as more variables are added.



Question 20.20. You run a correlation matrix between a Y variables auto sales in units and two X variables auto prices (X1) and car buyer’s income (X2). As expected auto prices had a high negative correlation to auto sales while buyer’s income had a high positive correlation. Both X variables had significant correlations. When you run a multiple regression analysis of the forecast variable auto sales with independent variables automobile price and car buyer’s income the results were positive coefficients for both price and income. Not only that the automobile price variable coefficient was found not to be significant. Is this what you would expect and what is the likely cause? (Points : 3) Yes, you would expect this as a result of serial correlation. Yes, you would expect this since you cannot tell regression model outcome from correlations. No, the switch in expected sign and lowered significance is likely caused by serial correlation. No. the lowered significance and sign switch is likely caused by multicollinearity.



Question 21.21. Why are heteroscedasticity, multicollinearity and serial correlation a problem in regression analysis? (Points : 3) They indicate that the regression model requires more data. They indicate that the model coefficient standard error is incorrect and their significance may be subject to Type I error. They indicate that the residuals are not normally distributed. Their presence indicates that the actual standard error of the model is too low and the model is not reliable resulting in Type 2 error.



Question 22.22. What is indicated when the actual future values of forecasted sales begin to fall outside of the forecast confidence interval? (Points : 3) The forecaster should collect more recent sales data and rerun or revise the regression model. The forecaster should adjust the forecast up or down depending on the direction of the new actual sales data. The forecaster should stick with the existing model since the data will likely begin to fall within the confidence interval. The forecaster should take a hard look at how the new sales data were collected since there is likely measurement error



Question 23.23. If the scatter plot between a dependent time series variable and an independent time series variable is curvilinear (curves upward over the data series) what should the forecaster investigate to improve the performance of a linear regression model for the variables? (Points : 2) Difference the X data to reduce its value relative to Y. Create a dummy variable that explains the curvilinear relationship. Seasonally adjust the Y variable to remove some of the variation. Use a natural log or other transformation of the X data as a dependent variable. Nothing, just run the regression on the data as it is since it cannot be improved.



Question 24.24. How do you calculate the Mean Squared Regression from the Sum of Squared Regression? (Points : 2) Take the square root of the Sum of Squared Regression. Divide the Sum of Squared Regression by the number of Y data observations. Divide the Sum of Squared Regression by the number of Y data observations minus 2. Divide the Sum of Squared Regression by the number of X variables used in the regression.



Question 25.25. One of the critical assumptions in regression is that the error term (et) has constant variance for all observations and Var(et) = s2. What problem in regression is caused if this assumption is violated and how is the problem detected? (Points : 3) Serial correlation and is indicated with the DW statistic. Multicollinearity and it is detected by the VIF. Heterscedasticity and it is detected by examination of the residual versus order or time series plot and applying the KB test. Non-correlation and it is detected by the significance of correlation coefficients. Non-normal residuals and it is detected by a histogram.



Question 26.26. You are responsible for forecasting company monthly product revenues for 4 months. You have collected the data found in Doc Sharing under Exam 3 Data. The X variables that you believe may determine sales revenues include customer use of the company website, per unit charge for the product, and the number of customers that are members of the company frequent buyers plan. In addition, you noted that sales revenues slumped during a worker\'s strike against the company in April of 2010.


Develop the best multiple regression model using this information and identify the appropriate R square value for the fit period below. Do not use transformations or a trend counter as X variables to get the best regression. (Points : 3) 97.0% 87.5% 98.8% 79.3% 99.5%



Question 27.27. Does the best regression model have significant multicollinearity? (Points : 3) No since the D-W statistics is below 2.5. No since the VIFs are below 2.5 Yes since the F value is very high at 745.9. No since the constant term is significant.



Question 28.28. Does the best regression model that you ran in problem 26 have severe serial correlation? How can you tell? (Points : 3) No since the D-W statistic is in the midrange between the lower limit and 4 minus the lower limit.. Yes, since the D-W statistic exceeds the lower limit. Yes, since the VIF is less than 2.5. No, since the F value and R square value are large. Yes, since the model D-W statistic falls below the D-W table value lower limit.



Question 29.29. Given the Hold Out (HO) values for each of the X variables and the fact that you are not expecting a worker\'s strike over the Hold Out period what is your forecast of sales revenues for August of 2011? (Points : 3) 144.9 168.2 158.8 152.8



Question 30.30. What is the MAPE for the forecast period for your best regression model? (Points : 3) 1.9 Percent 1.1 Percent 4.1 Percent 13.1 Percent



Question 31.31. Is the model heteroscedastic? Check the KB results to determine your answer (Points : 2) No the residuals do not indicate heteroscedasticity since the KB test coefficient is significant.


Yes, the residuals indicate heteroscedasticity since the KB coefficient is not significant.


Yes, the residuals indicate heteroscedasticity since the KB test coefficient is significant.


No the model is not heteroscedastic since the KB coefficient is not significant.




Question 32.32. The model accuracy was better for the forecast period than the fit period.(Points : 2) True False



Question 33.33. Does the hold out values fall within the forecast 95% confidence limits? (Points : 2) Yes. The hold out value all stay within the forecast confidence limits.


The hold out values fall outside of the forecast confidence limits in the last two months.


No. The hold out falls completely outside of the confidence interval for the entire forecast.




Question 34.34. While trying to move to incorporate quantitative forecast techniques in a company what is the best approach to address the existing executive level qualitative forecast method? (Points : 2) Use of a a multivariate regression model to demonstrate that it historically produces less error than the existing quantitative method.


Use of exponential smoothing to demonstrate that quantitative methods produce less error.


Use of regression to combine qualitative forecasts with quantitative forecasts to reduce error.


Run all four quantitative methods and produce a forecast with the lowest error and challenge the qualitative forecast results.




Question 35.35. What is a major indication that management has lost control over the operation of a company? (Points : 2) The lack of quantitative forecasting used for major business drivers.


Infrequent executive level meetings.


Fewer salary increases in key areas of the company.


Mistrust in the basic business driver forecast and forecast effort duplication.




Question 36.36. What is one of the most prevalent problems in business forecasting and what does it indicate? (Points : 2) Forecasts are not quantitative and it indicates a lack of management skill. Qualitative forecast methods are most times better than quantitative methods and more training should be targeted at qualitative forecasting. Executives don\'t trust each other and must rely on third parties to provide and unbiased forecast. The forecast function for the same data series is duplicated across the firm indicating a lack of strong leadership. Many people in the firm do not know how to use forecast information and, as a result, the firm is not efficient.


Paper#55638 | Written in 11-Dec-2015

Price : $37