Final Project Assignment17.871Spring 2002Assignment summaryYou will make two oral presentations, of 15 minutes in length, and turn in a final research paper,15–20 pages long.PresentationsGeneral considerations. Both presentations will be limited to fifteen minutes, followed by fiveminutes of response from me and others in the class. We will therefore start on time (at 5minutes past the hour) and I will cut you off when 15 minutes have elapsed, precisely. Fifteenminutes go faster than you imagine. The best presentations will be those that have been practicedbeforehand. Remember: you will be graded on the quality of your presentations. (Indeed, youwill be graded more on the presentation itself than on whether you actually found anythinginteresting in your research.) To improve presentation quality and to save time, you should havehandouts ready to distribute and/or have overheads ready to project. Make sure you have takencare of the technical aspects of the presentation before the class time begins.When you are not presenting, you are participating by listening and giving feedback. You mustbe present for all the presentation sessions.First presentation. The first presentation should inform us of your subject, how you intend topursue it, the data you plan to use, and any special problems you think you will encounter.Inform us of your subject: What is the substantive issue you intend to pursue? To whom wouldthe issue be of interest? Have others researched this area before? If so, what have they found?How do you intend to pursue the subject: What “unit of analysis” will you be studying? (That is,will you be studying individuals, counties, countries, etc.?) Will this be a time series analysis? Cross-sectional? A combination?Your data: What would the ideal data be to do this project? If you can’t get your ideal data, whatdata can you get your hands on? Tell us actual sources. It would also be good at this point toreport descriptive statistics of your data set, graphs that show simple, basic relationships, etc.Special problems. Perhaps there’s a crucial variable that will be difficult or impossible tomeasure directly, or some regression assumption will be violated. Quantitative analysis is rarelystraightforward, so tell us any wrinkles you’ve encountered thus far, or anticipate encounteringover the next month.2Second presentation. The second presentation should summarize the basics of your firstpresentation and then report your findings. The summary of the first presentation should be verybrief, focusing on reminding us of the subject you are pursuing, your basic approach, and thesource(s) of your data. Your findings should be reported in much the way we learned dataanalysis in this subject. First, present your data. Identify the variables in your analysis: how arethey measured, what are their means and dispersions? Second, present the most importantbivariate relationships. This may be done either with a correlation matrix or with a series of wellthought-out graphs. Third, present your multivariate analysis. Draw to our attention the coreresults and whether they confirm or disconfirm your orginal conjectures. Discuss any cases thatappear to be poorly described by your analysis and what might be done to correct this. Fourth,tell us what you conclude about your subject from the analysis you’ve done.Research paperThe logic of exposition in your research paper should roughly parallel the second presentation. The one section you should add at the beginning of the paper that you won’t want to talk toomuch about in the second presentation is a discussion of previous research. While you don’thave to do a comprehensive search of the literature, you need to do a little library work to see ifanyone has written about your topic before and (if they did) what they found. Don’t worry ifyou’ve found that someone has previously done something identical to your project. Replicationof others’ results is an important part of normal science.Journal articles in political science (which is the model your write-up should follow) often followthis outline:I. IntroductionSummarize your paper, providing the reader with a roadmap. Do not hide yourconclusions. Rather, state what your question is, why it's important, the generalstrategy you used to answer your question, what you found, and what theimplications of your findings are. Usually 2–3 pages.II. Literature reviewSummarize past research on the topic. Sometimes this will be research onprecisely topic you're writing on. Other times it will be research on a relatedtopic. The point is to place the current research into the context of past researchand the larger set of questions that research in this field has been pursuing. Usually 2–3 pagesIII. Overview of the empirical researchHow do you propose to address the broad question that you introduce in theintroduction? Usually this amounts to stating something to the effect that "inorder to understand X, we need to gather data Y and test to see if Z is true." Related to this, you should identify the basic moving parts of your empirical3analysis, including the unit of analysis and the independent and dependentvariables. This can usually be done in 2–3 pages.IV. DataWhere did you get your data and how did you measure you variables? Thisusually takes 1–2 pages. It is also usually helpful at this point to include a tablethat summarizes all your variables. Usually there are three (or four) columns: (1)concept, (2) indicator, (3) coding, and (4) actual data source. (The actual numberof columns may be less if, for instance, all the data come from the same datasource or if there's a particularly close correspondence between concepts andindicators [i.e., measurement ambiguity isn't really a problem].VI. ResultsA. Preliminary results (simple relationships and a first cut at multivariateanalysis)Start with the simple, straightforward analysis that is implied in SectionIII. Sometimes one or two simple bivariate graphs are useful to illustratethe more sophisticated multivariate analysis.B. Sensitivity analysis and other searches for anomaliesCheck to make sure that your results aren't driven by problems with data ormeasurement. For instance, if you believe that the result may be mostly aconsequence of a single outlying observation (e.g., you've included theDistrict of Columbia in a study of American states and you worry that DCis such an outlier that