Page 1C:\WP60\LECT2.PHD\FINALrev\Review05.Answers.wpdAnswers to the Review Questions for the Final Exam:Vartanian: Data II1. Cell 1: Father’s education has a more negative effect on the likelihood ofgrowing up poor relative to growing up middle class. Thatis, the higher the level of education, the less likely youare to be in category 1 relative to the excluded category.--Mother’s education has a more negative effect on growingup poor relative to growing up middle class. The higher thelevel of education of the mother, the less likely you are tobe in category 1 relative to the excluded category.--EDDADWF has more negative effect on growing up poorrelative to growing up middle class. The higher theeducation of the wife’s dad, the less likely you are to bein category 1 relative to the excluded category.–EDMOMWF has no effect.Cell 5: The higher the level of education of the dad, the morelikely the person will have grown up rich relative togrowing up middle class. The higher the value of EDDAD, themore likely you are to be in category 5 relative to theexcluded group.– The higher the level of education of the mom, the morelikely the person will have grown up rich relative togrowing up middle class. The higher the value of EDMOM, themore likely you are to be in category 5 relative to theexcluded group.–The higher the level of education of the father of thewife, the less likely the person grew up rich relative togrowing up middle class. The higher EDDADWF, the lesslikely you are to be in category 5 relative to the excludedcategory.– EDMOMWF has no effect on the outcome.From the second set of regressions, we can determine thedifferences between category 5 (Rich) and category 1 (Poor). Cell 5: The higher the level of education, the more likely youare to be in the rich group relative to the poor group.– The higher the level of mother’s education, the morelikely you are to be in category 5 relative to category 1. – The other variables are not related to differences incategory 5 and category 1. B. If the Hausman test showed significance, this would indicatethat we reject the null of iia, and we would therefore needPage 2C:\WP60\LECT2.PHD\FINALrev\Review05.Answers.wpdto run another type of test.#2. A. If we have a relatively large number of observations with 0 hours of work, our estimates for hours of work may bebiased. We run the selection model, examining factors thatmay affect whether or not someone works, and the correct forour estimates by this Heckman process. In other words, theHeckman selection model allows us to use information fromnon-working women to improve the estimates of the parametersin the regression model. The Heckman selection modelprovides consistent, asymptotically efficient estimates forall parameters in the model. b. There seem to be slight differences in the models. TheHeckman model seems to show that the effects of beingmarried are less than what the OLS model indicates. It alsoshows that the effects of owning your own home is morenegative than in the OLS model. The biggest differenceappears to be in the effects of health of the head. Theheckman models shows that the effects of having excellenthealth are quite a bit lower than the OLS model – anadditional 92 hours relative to those who don’t haveexcellent health, versus 104 in the OLS model for those withexcellent health relative to those with less than excellenthealth.C. This tests to determine if the error terms in the two modelsare related to each other. If they are not, then the OLSregression estimates are unbiased. If they are related toone another then the OLS regression estimates are biased. The null hypothesis is that Rho =0. Here, we reject thenull hypothesis and state that Rho…0, in all likelihood, andwe have better estimates with our selection model.3.A.Page 3C:\WP60\LECT2.PHD\FINALrev\Review05.Answers.wpdIn other words, there is a 31% chance that someone will buy acar.B. Thus, with these characteristics, the probability of buying a carincreases from 31% to 77%.C. In other words, people with 3 kids are .0578/1 as likely to buy acar as those people with 0 kids. Or, people with 0 kids are1/(0.0578) = 17.30 as likely to buy a car as those people with 3kids.D. Females are 4.48 times as likely to buy a car as males, controlling for the other variables within the model.5. A. Logistic Regression is your best choice.B. DV=likelihood of drug use IV=Peer pressure, either measured at the interval or nominal level. Control=Family income, district income and whether or not the person smokes.Page 4C:\WP60\LECT2.PHD\FINALrev\Review05.Answers.wpdC. The null hypothesis is that peer pressure will have no effect on the likelihood of drug use.D. You'll accept the null hypothesis in the first model where peer pressure is measured at the nominal level. In thesecond model, where peer pressure is measured at theinterval level, you'll reject the null hypothesis at the .05level.E. For people who experience peer pressure, have $0 in family anddistrict income, and do not smoke, the probability of druguse isThe probability for those people who do not experience peerpressure but who smoke is5.A and B. Children with parents who drink a lot of caffeine are found to be more likely to throw rocks than children with parents whodrink little or no caffeine. You will thus accept your nullhypothesis that there is no relationship between caffeinedrinking parents and rock throwing since the relationship isin the opposite direction of your research hypothesis. Those children who ride motor scooters to school are shownto be less likely to throw rocks at birds. In fact, this isPage 5C:\WP60\LECT2.PHD\FINALrev\Review05.Answers.wpdthe best predictor we have of whether or not a child willthrow rocks at birds. Children who ride motor scooters areonly 1.8% as likely to throw rocks are children who do notride motor scooters. This means that children who do notride motor scooters are 54.598 times as likely to throwrocks as children who ride motor scooters (1/0.0183). Aspredicted, Libras are more likely to throw rocks, butunexpectedly, Scorpios are less likely to throw rocks thannon-Scorpios. VariableNull Hypothesis LikelihoodCaffeine: Accept 49% more likely for those who drink 1 ounce versus those who drink 0ounces of caffeine. Scorpio: Accept 13.53% as likelyLibra: Reject 7.389 times as likelyScooter: Reject 1.83% as likelyC. Yes. You've made a type II error. This is when you accept thenull hypothesis