Graduate level Statistic writing assignment. MUST be 3-to-5-page paper (double-spaced, not including tables and graphs) MUST know how to use STATA.
PAPER ASSIGNMENT 1 Due date: Monday, October 18, 6:00 a.m. Describe the impact of at least four independent variables on an appropriate dependent variable of your choice in a 3-to-5-page paper (double-spaced, not including tables and graphs). You can describe current public opinion (e.g., using the 2018 General Social Survey) or trends in public opinion (e.g., using the General Social Survey 1972-2018 cumulative data file). You can also use one of the ACS data sets to look at earnings, hours worked, or managerial status. Use this paper as an opportunity to show me how much you have learned in this class. A variety of data sets are posted on the Stata Data Sets page in iCollege. If you have your own data set, please get my permission before using it. If you have a particular topic that interests you but you don’t have a data set, please check with me to see if I have ideas. Choose an interesting dependent variable. (Please try to make it interesting for me, too. I have read dozens of writing assignments about capital punishment, marijuana legalization, abortion rights, and earnings. If you want to choose one of those, please try to be original in other ways.) Choose at least four independent variables that you expect to have a causal impact on your dependent variable. Be sure you understand how the variables are coded. Use Stata commands (e.g., describe, codebook, label list) to see what variables are in the data set and how they are coded. Use the GSS Data Explorer to get the exact question wording if you are using one of the GSS data sets. Also use Stata to get rid of missing values and to make sure your variables are coded in the ways you want. If you are examining public opinion, you should probably make your dependent variable a dummy coded either 1 or 100 for Yes and 0 for No. (The 0-100 coding works better for bar charts and other methods where you want the percentages who said Yes. If the original variable is ordinal level, present the frequency distribution for the original coding and explain how and why you dichotomized it. If the original variable is coded “.” you should keep it as a missing value, as the respondent was probably not asked the question. If the original value is “Don’t know” or “Know answer,” you can recode it either to missing (“.” ) or to zero. If you make it missing, you are reporting the percentage of people who expressed an opinion If you make it zero, you are reporting the percentage of people who were asked the question. Feel free to transform your variables to make them more useful. If you have a good reason, recode ordinal-level into dummy variables or interval-level into ordinal-level variables. Justify your recoding. Write up a 3-to-5-page report (plus tables and graphs) using the following format: Title Your title should at least identify the topic you are examining. Ideally, it should convey your research question and perhaps even be clever. Introduction: Write 2-4 sentences making clear what research question your paper is trying to answer. Your research question is almost certainly about what factors influence your dependent variable (e.g., “What personal characteristics affect whether people think torture of suspected terrorists is at least sometimes justified?”). Briefly explain why your question is worth answering. Hypotheses: State and justify three hypotheses about factors that influence your dependent variable – that is, state how you think your independent variables will affect your dependent variable and why you think that (1-3 sentences each). Word hypotheses differently for dummy and interval-level independent variables. For dummy dependent variables, talk about probabilities. Say something like “Men will be more likely than women to …” or “As conservatism rises, the probability of saying/doing … will increase/decrease.” For interval-level dependent variables, talk about expected values. Say something like “Men have lower/higher salaries/scores/etc. than women …” or “As conservatism rises, expected salaries/scores/etc. will increase/decrease.” For dummy and nominal-level independent variables, make comparisons. Say something like “Men will be more likely than women to …” or “Asians will have higher average salaries than whites.” For ordinal-level independent variables, talk about how the dependent variable will change as the independent variable increases. Say something like “As conservatism rises, people will be more likely to …” or “As education rises, salaries will tend to increase.” All hypotheses should be causal in nature. That is, you expect that a change or difference in your independent variable will lead to a change or difference in your dependent variable. Remember that causes precede effects in time. Justifications can be very brief – they just need to show why you expect your hypothesis to be true (e.g., “Men will be more likely than comparable women to believe that torture of suspected terrorists is at least sometimes justified. This is because men are more violent than women”). Data: Very briefly describe the data set: who gathered the data, when, how, and how large the sample size is. State how you operationalize (code) your variables. Typically, start with the dependent variable, which will almost certainly require giving the exact question wording. Tell us how you code it. If you are going to use bar charts or line graphs, you should almost certainly code it 100 for Yes and 0 for No. Yes could be either Favor or Oppose, depending on how you want to focus your paper. As a reader, I do not need to know what the original coding was – that is, I do not need to know the names of the original or the new variable or what values the original variable had. I just need to know how the variable you are analyzing is coded. (For instance, “The dependent variable is coded 100 for those who think we are spending too much on welfare, and 0 for those who think we are spending too little or about the right amount.”) As an instructor, I want to be sure you recoded your variables correctly. Leave the original variable coding out of the text but add an appendix with your do-file, including the commands you ran to create your new variable, plus the command that ran a crosstab of the original and the recoded variable to double-check your work. Many independent variables will be clear enough that you do not need to provide the question wording, but add it if necessary. Use clear value labels for all variables. Do not include crosstabs that show values of 100, 1, or 0; make sure your reader understands what all the rows and columns mean. Methods: Explain how you are going to test your hypotheses. You can use crosstabs, bar charts, or line graphs to look at differences in percentages. You can use tab indvar, sum(depvar), bar charts, or line graphs to look at differences in means. You should probably incorporate prtest or ttest into your analysis to be able to generalize to the broader population. (Be clear what population you are generalizing to.) If you feel up to it, include a second or third independent variable in your analysis. You can do that with crosstabs, bar charts (remember asyvar), or line graphs. Findings: Begin by describing the overall pattern for your dependent variable – e.g., the percentage of the full sample that say Yes (or something else) or the mean earnings of employees in a particular sector. Think about adding a 95% confidence interval here. If you are looking at opinion change over time, describe the overall trend next. Present your tables and findings in the same order that you present your hypotheses. You probably want to present findings for your independent variables in the same order that you mentioned them in your hypotheses. Many authors number their hypotheses, which helps with ordering. You should probably present your tables or graphs in the same order. Number them and refer to them by number in the text. Remember that I usually do not need to know what percentage of the sample has particular values of the independent variables. What I need is to know what percentage of those with each value of the independent variable say Yes. Think about adding confidence intervals and hypothesis tests: To what extent can you generalize from the sample to the population? Remember to describe differences in probabilities in percentage points, rather than %. Tables and graphs: Insert your tables and graphs either in the text very close to where you discuss them or at the end of the paper, before the appendix. Give them numbers and refer to them by number in the text. Do not hide your tables in the appendix so that I have to go looking for them. Make your tables, bar charts, and/or line graphs as attractive as possible. Ideally, your reader will be able to understand your story without reading your text. Add variable and value labels that make sense and look good. Only ask for the percentages you need in your crosstabs; leave out the absolute frequencies in most cases. You can probably just report what you need from prtest rather than including the output. Try copying the Stata output into Excel, throwing out all the rows or columns you don’t need, and getting rid of excessive digits before you copy into Word. Give titles to your graphs. Make sure you label the X- and Y-axes clearly. Can you use color and line width or patterns better? Style: Explain your findings in simple, straightforward language that someone not paid to do so would be willing to read. Do not repeat the same paragraph over and over, just changing the variables and numbers. Try to make it interesting to read. Throughout your paper, strive to use active voice verbs. Avoid the passive voice. Try not to start sentences with “There is.” If, when you are editing your paper, you notice that you have started sentences with phrases like “It could be argued that” or “It is believed that” or “Extant research indicates that,” delete everything up through the “that.” Look back at the sentences and phrases I have crossed out in your writing assignments. Avoid putting similar sentences and phrases in your paper. You might look at how Gallup and the Pew Research Center write up their findings. Here are a couple of trend analyses: https://news.gallup.com/poll/267698/support-legal-marijuana-steady-past-year.aspx and https://www.pewforum.org/fact-sheet/changing-attitudes-on-gay-marriage/ Check their websites for discussions of topics more similar to your dependent variable. Final tip: If you get surprising findings (e.g., that men were more likely than women to support Hillary Clinton), check your work to be sure that you did not make mistakes. Appendix: Include all your commands. You might want to add some output that you did not put in the tables section; for instance, if you rely primarily on bar charts or line graphs but want to refer to some numbers in the text, you might include the tables that generate those numbers in the appendix.