Statistics Immersion Program - Seminar in Quantitative Analysis | POLI 7962, Study notes of Political Science

Download Statistics Immersion Program - Seminar in Quantitative Analysis | POLI 7962 and more Study notes Political Science in PDF only on Docsity! Statistics Immersion Program POLI 7962: Seminar in Quantitative Analysis Summer 2008 MTWTh 8:30 – 11:30 Stubbs 220 James Garand Emogene Pliner Distinguished Professor Stubbs 205 Office Phone: 578-2548 Email: pogara@lsu.edu Web site: http://jgarand.lsu.edu/ Torture numbers, and they'll confess to anything. --Anonymous cynic And if California slides into the ocean, like the mystics and statistics say it will, I predict this hotel will be standing until I pay my bill. --Warren Zevon, from "Desperadoes Under the Eaves" Introduction The purpose of this course is to introduce students to a range of basic statistical and data analytic techniques necessary to understand and conduct quantitative social, political, and policy research. The development of such methodological skills is an absolute necessity for social scientists. Professional social scientists are often called upon to either conduct quantitative research on their own or, at the very least, be able to understand, interpret, and utilize the considerable body of social research that employs quantitative analytical techniques. Courses in statistics subsequently should not be viewed as an obscure degree requirement for the M.A. or Ph.D. programs in social science disciplines, but should instead be seen as providing an introduction to the requisite research skills for practicing social scientists. While the topics covered in this seminar cannot in any way be considered exhaustive, they do represent many of the basic statistical issues with which social scientists should be familiar. Several topics will be examined in this course. First, we will discuss briefly the philosophy of social science and the role of quantitative methods in conducting research in the social sciences. Second, we will discuss various statistical techniques utilized in univariate analysis--i.e., measures of central tendency and dispersion. Third, the logic of statistical inference and hypothesis testing will be examined. Of primary interest in this section will be the estimation of population parameters (characteristics) based on information collected from random samples drawn from populations. Finally, various bivariate and multivariate statistical techniques will be discussed. Because social science often focuses upon the relationship between two (or more) variables, special emphasis will be placed on establishing the magnitude and direction of such relationships as they exist within both populations and samples of populations. Two major points should be made about this course. First, one of the best ways to learn about statistical techniques is to practice them as much as possible. Hence, on most days each student will work several problems 2 that require the use of the various statistical techniques examined in this course. By going through the process of computing the answers to statistical problems, it is hoped that each student will develop the statistical skills necessary to understand and conduct empirical research. Second, one major goal is to help students identify and discard numerous myths which pertain to statistical analysis. To take one example, many students in the social sciences adhere to the view that it is easy "to lie with statistics." While there is some truth to this proposition, for the most part lying with statistics is a more successful strategy when one lies to individuals who do not understand statistical methods. (I would revise this "truism" to state that it is easy to lie with statistics to those who don't understand statistics!) Social scientists who are well-trained in statistical methods can usually differentiate good statistical arguments from bad ones. One goal of this course is to give social science students (and others) the skills necessary to analyze statistical arguments and the validity of inferences drawn from statistical statements. Graduate Assistants: There are two graduate assistants for this program. The graduate assistants will direct the laboratory sessions for this course, and they will be available for consultation. Graduate Assistant Email Address Office Number Betina Wilkinson bcutai1@lsu.edu 318 Stubbs Hall N. Kim Nguyen nnguy35@lsu.edu 307 Stubbs Hall Reading The following book has been ordered and is available at the University bookstore: David Knoke, George W. Bohrnstedt, and Alisa Potter Mee, Statistics for Social Data Analysis (4th edition) (ISBN 0-87581-448-4) Companion Course This program also includes a required laboratory course. POLI 7964 will be directed by Betina Wilkinson and Kim Nguyen and will be offered from 1:30 to 4:30 each afternoon in Middleton 232. The primary purpose of the lab is to introduce students to computer applications appropriate for statistical analysis and to follow up on issues raised in the morning seminar. The following book has been ordered for the laboratory sessions and is available at the University bookstore: Philip H. Pollock III, An SPSS Companion to Political Analysis (3nd edition) (ISBN 978-0-87289-607-9) Course Web Site I have created a web site for this course. The site will include assignments, data sets, links to statistics web sites, and other helpful information. The course web site can be found on my personal homepage at: http://jgarand.lsu.edu/ 5 Course Outline Week 1 July 28 Introduction: The Role of Statistics in Political and Social Research Variables, Data, and Measurement Knoke et al., Chapter 1 July 29 Univariate Analysis: Frequency Distributions Knoke et al., Chapter 2, pp. 29-40 July 30 Univariate Analysis: Central Tendency and Dispersion Knoke et al., Chapter 2, pp. 40-63 July 31 Sample Estimation of Population Parameters Knoke et al., Chapter 3. Week 2 August 4 Sample Estimation of Population Parameters (continued) August 5 Bivariate Analysis: Difference in Two Means Knoke et al., Chapters 4-5 August 6 Bivariate Analysis: Analysis of Variance (ANOVA) Knoke et al., Chapter 4 August 7 Bivariate Analysis: OLS Regression Knoke et al., Chapter 5 (skim) Knoke et al., Chapter 6 Week 3 August 11 Bivariate Analysis: OLS Regression (continued) August 12 Bivariate Analysis: OLS Regression (continued) August 13 Multivariate Analysis: Multiple Regression Knoke et al., Chapter 7-8 August 14 Multivariate Analysis: Multiple Regression (continued) 6 An Aside on the Development of Methodologists By James E. Campbell SUNY at Buffalo It has been my observation that there are five phases or stages in the development of orientations toward quantitative methods by political scientists. It may be helpful to some of you to examine these stages and recognize your own stage of development. 1. The Know-Nothing Stage The Know-Nothing Stage is characterized by an absence of knowledge of statistical techniques and the quantitative approach, as well as a companion and compensating defensive resentment toward these methods. Fear of the unknown causes some to dismiss the importance of the unknown. If I don't know or understand it, it can't be important. Quantitative analysis is just "numbers crunching" or can't be brought to bear on the really important political questions. From this perspective, ignorance is not only bliss; rather, it is somehow intellectually superior to the precision and comparability offered by a more rigorous kind of empirical method. 2. The Novice Stage This is a giant step removed from the Know-Nothing Stage. The Novice recognizes, at least theoretically, the value of quantitative analysis. What is distinctive about the Novice Stage is uncertainty and tentativeness. The Novice is usually unsure of the methods used, unsure of their properties, and even more unclear about how they can be applied to political questions. This is the stage of basic learning, both in the sense of learning by the statistical book and learning by becoming comfortable with statistics as a tool and seeing its applicability to various political questions. 3. The Mad-Correlator Stage Also known as "Barefoot Empiricism." These are the guys who give statistical analysis a bad name. They are the analysts who have learned to use statistical techniques and are hell-bent on using them indiscriminately on any data set on which they can get their grubby little hands. They know statistical analysis only to a "cookbook" degree. Their understanding is superficial in two very important ways. First, they emphasize the use of the technique over any possible limitations of that technique. Second, they fail to understand that quantitative methods are just that--a method or a tool. To be used properly a tool must be used with a plan to guide it--in the case of social science, the plan is based on theory. It is common at this stage of development for one to see no association between statistical methodology and the course on the philosophy of social science that the Mad Correlator was forced to take. Understanding the connection between the two is necessary to advance to the next stages of development. 4. The Purist Stage At the Purist Stage the analyst is not only competent with the use of quantitative analysis, but he or she is also acutely aware of the assumptions and limitations of the techniques. These folks play the statistical game strictly by the book. These are the folks who insist on interval-level data before using parametric statistics. For the Purist, a violated assumption carries a stiff penalty. They regard statistical techniques as tools but treat them as fairly fragile tools. They are right to recognize violated assumptions or other problems where they exist, but they are often deficient in failing to distinguish between important and unimportant problems. They know how the tool ought to be used but they have allowed this to get in the 7 way of getting the job done. In a way the Purist is captured by the tool or statistical technique. He or she is so intent on using the technique properly that the purpose for using the technique in the first place gets second billing. 5. The Sophisticated Stage This is the stage that I hope all will be in by the end of this course, or at least by the end of the second semester statistics seminar. The Sophisticated Analyst knows the uses and limitations of quantitative methods thoroughly. He or she is guided in the use of these techniques by a general concern for theory building and a specific concern for theories in the particular area of inquiry. The Sophisticated Analyst is sensitive to the limits of techniques, but is not hypersensitive. He or she understands that some technique problems make no substantive difference while others may raise serious doubt about conclusions. The Sophisticated Analyst examines analyses from this perspective and may examine data in a number of different ways to check for the reliability of his or her results. Comparing the Mad Correlator, the Purist, and the Sophisticated Analyst: The Mad Correlator is too aggressive with data. He or she runs the risk of reading more into the data than is actually there. The Purist, on the other hand, is too timid. The Purist gets less from the data than there is. The Sophisticated Analyst gets as much from the data as is there--no more, no less.