ECN 410 PRACTICE EXAM

Referred to output and your knowledge in order to answer the question: least squares estimates are still blue - correct ans-false Density and income are not included in the model, they become correlated error, increasing se of the regression (variance) and the coefficients become biased Consider the following fitted regression model: Square root miles = 8.24 + 0.23 income + 0.24 age - 1.13 kids The correct interpretation is: an increase of $1000 in income, will increase on average, square root of 230 miles or 15.16 miles, holding everything else constant - correct ans true Straight from the slides (endogeneity lecture) In the study of 28 industrial establishments of of varying size, the number of supervised workers x and the number of supervisors y were recorded. A regression model was developed to investigate the relationship between y and x. Use output c to answer the question: which classical assumption was violated? 1) the regression model is linear is correctly specified and has a additive error term 2) the error term has zero population mean 3) the error term has a constant various (homogeneous variance) 4) the error term is normally distributed 5) all explanatory variables are uncorrelated with the error term (no endogeneity) - correct ans-3) the error term has a constant variance ( homogenous variance) 1) (no violation of ca. The lack of fit test showed a linear relationship.) 2) (no violation of ca. The sas output shows the mean of residuals = 0) 3) (true. The hypothesis of constant variance was rejected.) 4) (no violation of ca. Normality tests showed p-values 0.05.) 5) (true, even though there is no correlation output, the data contains only y and x, therefore, including x in the regression model eliminates any specification bias.) Referred to output a and your knowledge in order to answer the question: all the explanatory variables are uncorrelated with the error term (use a = .10) - correct ans false Both income and density are correlated with the error term We would always like to reject the null hypothesis when testing for normality of the errors - correct ans-false Use output c to answer the question about the residuals: ECN 410 ECN 410 1) the plot are the residuals versus x shows that the residual variance tends to decrease with x 2) the residuals tend to lie in a band that diverges as one of those along the x-axis 3) because the band within which the residuals lie diverges as x increases the error variance is also increasing with x 4) a plot of the residuals against the predictor variable points up of the the presence of heteroskedastity errors 5) evidence that we need to transform the dependent variable to fix the regression model - correct ans-only 1 is false (the plot of the residuals versus x shows that the residual variance tends to increase with x) When a researcher leaves important independent variables out of the regression equation, the estimates are not blue anymore - correct ans-true Using sas output b which of the classical assumptions were tested 1) multicollinearity 2) normality of the errors 3) linearity 4) e=0 - correct ans-only 1 is false Blue means best linear untransformed estimators - correct ans-false When testing for normality, sas output give 3 goodness of fit tests. The decision is based on when the majority reach agreement - correct ans-true When a researcher leaves important independent variable(s) out of the regression equation he or she violates the following classical assumption: Observations of the error term are uncorrelated with each other - correct ans-true Which of the following is used to test for heteroscedasticity? 1) anderson-darling test 2) heteroscedasticity-correction test 3) kolmogorov-smirnov test 4) breusch-pagan test - correct ans-only 2 and 4 are true Referred to output a and your knowledge in order to answer the question: The least square estimates are unbiased - correct ans-false How to detect heteroscedasticity in a regression model: 1) look at the residual plots against each independent predictor. V or u shaped pattern indecates that the error terms do not have homogeneous variance 2) check if the regression models is linear and there is no limited variables 3) test equality of variances using the lack of fit test ECN 410 ECN 410 4) using the spec option in sas, if you fail to reject the no hypothesis then we can conclude that there is heteroscedasticity in the regression model - correct ans-only 1 and 2 are true (iii is false, lack of fit test for linearity not heteroscedasticity. Iv is false, if we fail to reject the null hypothesis, the variances of the error terms are constant.) If the equation in a model has a log transformaton the interpretation of the slope should be avoided - correct ans-false Refer to output a and your knowledge in order to answer the question: I. The error term is normally distributed. Ii. The are no outliers in this data that can compromise the normality of the error terms. Iii. The probability of not rejecting the null hypothesis when testing for nomality of residuals is higher than 10%. Iv. The sample data is in accordance with the central limit theorem and non normality it is not an issue for this regression model. All are true. I, ii and iii are false. I, iii and iv are false. All are false. - correct ans-all are true. I. True, all normality test indicate a p-value 0.10 (see table below) Ii. The are no outliers in this data that can compromise the normality of the error terms. True, there is no issue of non-normality, so outliers do not compromise the normality of the errors. Iii. The probability of not rejecting the null hypothesis when testing for normality of residuals is higher than 10%. True, all p-value are greater than 0.10 or 10% Iv. The sample data is in accordance with the central limit theorem and non normality it is not an issue for this regression model. True, n = 51 therefore 30, clt stands for this data. Consider the following fitted regression model: calls = -0.14 + 2.31 months - 0.04 months^2 , where calls are number of calls placed per day, and months are time on the job. The correct interpretation is: an additional month on the job will increase the number of calls by (2.31 - 2*0.04) = 2.23 True False - correct ans-true If x increases by one unit, y will change by (b1+b2x) units (2.31 - 2*0.004) = 2.23 Using sas output b, choose the correct answer: - the log-log transformations took care of the outliers in the data. ECN 410 ECN 410 - the standard deviation of the errors in model 2 can be calculated taking the natural log of the standard deviation of the errors in model 1. - because the ss pure error went down in model 2, we can affirm that log-log transformation took care of the non-linearity issue. - the residuals are not normally distributed in both models because n 30. - correct ans-- because the ss pure error went down in model 2, we can affirm that log-log transformation took care of the non-linearity issue Using sas output b, which of the sas codes does not correspond to the output shown: 1) proc reg data = eggs; Model sales = price_eggs/vir; Output out = resid r = res; Run; 2) data one; Set eggs; Insales = log(sales); Lnprice_eggs = log(price_eggs); Run; 3) proc reg data = one; Model sales = price eggs; Model insales = lnprice_eggs/lackfit; Output out = resid1 r = res; Run;quit; 4) proc univariate data = resid1; Var res; Histogram/nornal; Run; 5) proc gplot data = eggs; Plot sales*price_eggs; Run; I and v are false. Only i is true. Only i and iii are true. Only i is false. - correct ans-i and v are false I. Output b does not show variance inflation fa