This document provides instructions for conducting several logistic regression analyses using different datasets and interpreting the results. The analyses predict views on prison sentencing severity using ideology and gender (exercise 1), support for abortion based on religiosity, political views, premarital sex attitudes, and gender (exercise 2), and compare likelihood ratio and z-tests of predictors (exercise 4). It also asks the reader to conduct a power analysis to determine sample size needed to detect an effect of education on views of sentencing when controlling for other variables (exercise 5).
This document provides instructions for conducting several logistic regression analyses using different datasets and interpreting the results. The analyses predict views on prison sentencing severity using ideology and gender (exercise 1), support for abortion based on religiosity, political views, premarital sex attitudes, and gender (exercise 2), and compare likelihood ratio and z-tests of predictors (exercise 4). It also asks the reader to conduct a power analysis to determine sample size needed to detect an effect of education on views of sentencing when controlling for other variables (exercise 5).
This document provides instructions for conducting several logistic regression analyses using different datasets and interpreting the results. The analyses predict views on prison sentencing severity using ideology and gender (exercise 1), support for abortion based on religiosity, political views, premarital sex attitudes, and gender (exercise 2), and compare likelihood ratio and z-tests of predictors (exercise 4). It also asks the reader to conduct a power analysis to determine sample size needed to detect an effect of education on views of sentencing when controlling for other variables (exercise 5).
the person sees prison sentences as too severe (coded 1) or not too severe (coded 0). The variable liberal is how liberal the person is, varying from 1 for conservative to 5 for liberal. The variable female is coded 1 for women and 0 for men. Do a logistic regression analysis with severity as the dependent variable and liberal and female as the independent variables. Carefully interpret your results. 2. Use gss2002 chapter11.dta. What predicts who will support abortion for any reason? Recode abort12 to be a dummy variable called abort. Use the following variables as predictors: reliten, polviews, premarsx, and sei. You may want to create new variables for these so that a higher score goes with the name; for example, polviews might be renamed conservative because a higher score means more conservative. Do a logistic regression followed by the listcoef command. Carefully interpret the results. 3. Using the results from exercise 2, create a bar graph showing the percentage change in odds of supporting abortion for any reason associated with each predictor. Justify using a 1-unit change or a 1-standard-deviation change in the score on each predictor. 4. Using the results from the logistic regression in exercise 2, compute likelihoodratio chi-squared tests for each predictor, and compare them with the standard z tests in the logistic regression results. 5. In exercise 1, you did a logistic regression using gender and how liberal a person was to predict whether they thought prison sentences were too harsh. Suppose that you think people with more education are likely to feel that sentences are too severe. Do a tabulation on severity to see what proportion overall think sentences are too severe. Then do a power analysis to see how big a sample you would need to have a power of 0.80 to detect whether a 1-standard-deviation increase in education would raise the proportion of people who think sentences are too severe by 0.10 when liberalism and gender are already in the model. You would be adding education as a third predictor. State the information you need and the values you assume. What assumption of the command you used did you violate?