The following are things to think about when you submit a research paper (term paper, preliminary PhD paper, dissertation proposal, comprehensive exam, journal submission).
"No-Brainers"
2. Run a grammar-check program
3. Spell-check and grammatic check programs are not foolproof.
Double check your work to make sure you use words appropriately, that your sentences make sense, that you don't repeat yourself. In particular, be careful how you use the following words:
4. Look at your paper after you print it (and before you turn it in).
Everything may look fine on the computer screen but the printer may not work correctly, producing strange characters or simply run out of toner. Check to see if page numbers did appear on each page. Make sure no one else picked up your paper first and accidentally resorted the pages. Make sure the tables and figures aren't stapled upside down at the back of the text.
5. Start your project on time.
Work on a semester paper should not begin two weeks before the end of the term. Many of the problems noted in the papers this semester are the result of a late start in data collection, data cleaning, and preliminary analysis. See below.
6. Always keep backup copies (paper and diskette).
Quantitative issues
1. Examine your data. "Clean" your data.
Before you begin to work seriously on any statistical analyses, you have to become familiar with your data. That means you have to examine it in great detail. Run frequencies and compute basic descriptive statistics on each "raw" variable (and on each new variable you create from existing variables). What is the lowest value? What is the highest value? Are these reasonable? Are missing data indicators set correctly? A second check can be to graph your data. Do you get strange "outliers" when you look at your data?
If the above examination indicates problems, you have to fix them. This is what I mean by CLEAN THE DATA. Missing data and typos can destroy analyses of any data set and are particularly effective in reducing the usefulness of small N data sets.
DO NOT BEGIN TESTING YOUR MODEL UNTIL YOU HAVE EXAMINED THE DATA. DO NOT STOP CHECKING YOUR MODEL FOR ODD RESULTS ONCE YOU BEGIN.
2. Measurement, measurement, measurement
I'm not convinced that you can spend too much time on your data sources, your measures of key variables, and how you operationalize key concepts. You cannot expect a reader to follow your analysis of democracy or public opinion (broad concepts usually measured with narrow quantitative variables) if he or she does not follow your measurement strategies. Why do you measure X this way? Why is this measure preferred? Any good research paper will carefully walk through the measurement process.
3. Univariate Statistics. (and N's)
Many of you included a table of univariate statistics representing the maximum and minimum, mean and standard deviation (or variance) of your variables. This is a natural extension of the previous discussion of measurment and I was pleased to see these tables. They provide easy introduction to the data for the reader. But what was missing from each was the valid N's for these variables. It can provide some considerable insight into why many of your presented extended analyses of a much smaller subset of your data than, for example, the number of respondents surveyed.
4. Bivariate Statistics (aka correlation matrices)
Just as univariate statistics offer some basic insights into the state of your data, bivariate statistics provide an initial sense of relationships among variables. Unless you're analyzing an unwieldy number of variables (e.g., the Baldus study referenced on the final exam), consider presenting either a complete correlation matrix or small matrices relevant to subsets of your model such as relationships among items combined in a scale.
5. Do not use variable names in the body of the paper or in tables.
Just because the variable LOGMC1 means something to me, doesn't mean that you (or anyone else) will know what concept it is measuring. Always refer to concepts over measures or variable names--after all, that is what your theory is built upon, right?
Stylistic issues
1. Avoid down-playing your analysis or over-emphasizing everybody else's.
This is generally a problem in Graduate School, but one which we all need to grow out of. Literature reviews of the "he said/she said" type are boring to read. Quickly get to the point and showcase your theory, your argument, your analysis, and your conclusions.
2. Highlight your contribution.
What are you adding to the existing body of knowledge in Political Science? This is the purpose of much of the written work that we do as social scientists.
3. Shorter papers read better.
Frequently, you can get your point across in fewer pages. Always try to find ways to eliminate tangents, unrelated commentary, redundancies, and plain old "fluff" from your written work. The best papers are those which are thorough but concise--nobody likes wading through 50 pages of redundancies and theories which are not even explored in the analysis section of your paper.
RETURN to Term Paper Assignment.