a. Test the regression into 3 model: log, non-linearity, quadratic model : birthweight= Bo+ B1 Drink + B2 NPRE visit + B3 Age + B4 Smoke + B5 Alcohol. Which is the estimated regression equation that best fits your data? b. Next, you will test the 4 assumptions of the classical linear regression model. testing: (i) Normality of errors (2) (ii) Multicollinearity (2) (iii) Heteroscedasticity (for rigors, use at least 2 tests) (3) (iv) Autocorrelation (for rigors, use at least 2 tests) (3)

Steps: a. State the regression equation you selected from Task #2. b. Test for normality of errors and interpret your results. c. Test for multicollinearity and interpret your results. Does your test suggest that multicollinearity may be causing a problem in your estimation? If so what remedies will you apply? After applying the remedies, run your new regression, report and interpret your new regression. If you determine that multicollinearity may not be causing any problems, then move on with your original regression d. Based on the regression in (c) above, perform at least 2 tests for heteroscedasticity. Interpret your test results. Do your tests suggest that the BLUE qualities of OLS may be compromised? If so, fix the heteroscedastic errors by running a regression on (c) above with heteroscedastic corrected errors. If the BLUE qualities of OLS are not compromised, then move on with your regression in (c). e. Based on your regression in (d), run at least two test for serial correlation. Interpret your test results. Do your tests suggest that BLUE qualities of OLS may be compromised? If so, fix the autocorrelated errors by running a regression on (d) above with autocorrelated corrected errors. If the BLUE qualities of OLS are not compromised, then move on with your regression in (d). f. Present your final (corrected) regression model with all the appropriate interpretations and diagnostic tests. (t-test and F-test) Your final report should be no more than a page and half excluding the appendices.