category: Statistics

Most academic and policy research focuses on explaining variation in a dependent variable, asking what causes some people to act or believe one way

and others to act or believe differently.  For example, what demographic characteristics and attitudes of respondents influence their salaries,

political tolerance, political ideology, or opinions on government spending?  Use one of the class data sets or another data set, if you have one, to

write a brief research paper trying to explain an interval-level dependent variable.

You should begin your paper with a statement of the research question and a brief justification for investigating that question, probably based

on why the issue is theoretically or practically important or interesting.  You should then state your ideas (hypotheses) about relationships among

the variables you are looking at.  You should justify these hypotheses logically, based on theoretical reasoning, research done by others, or common

sense.  You will probably use several independent variables in a multiple regression analysis.  Next, you should briefly describe the data set.  As

appropriate, the body of the paper should address the following questions:  what is the distribution of your dependent variable?  is your dependent

variable related to the other variables in the directions you hypothesized (perhaps in both bivariate and multivariate analyses)?  which

relationships are strongest/weakest?  which relationships are statistically significant?  is some of the bivariate relationship between variables

spurious or indirect?  The paper should close with a brief conclusion that summarizes the judgments you have reached and possibly draws some

additional implications.

Although some reading on your research topic would not hurt, the emphasis in this paper is on analyzing the data and describing your analyses.

You should show your ability to develop meaningful regression analyses, to interpret the appropriate statistics,  and to write sensibly about your

analyses.  Your conclusions should be based on the data and you should turn frequently to the data in supporting your points in the text.  The

emphasis should not be on how you calculated the statistics but on what they mean.  For instance, don’t do a formal, five-step hypothesis test.

Instead, indicate whether each relationship is statistically significant (probably with a brief statement of how you know) and what we learn from

knowing whether the relationship is statistically significant or not.

Your paper should be concise, grammatical, carefully edited for spelling errors, about three but definitely no more than five double-spaced pages

long (with one-inch margins and 12 point font), and be typed or written on a word processor.  You may add up to five pages of graphs and tables to

your five pages of text.  Tables may be written by hand, typed up on a word processor, or pasted from SPSS.

Attach your output as an appendix to the paper.  I will not refer to the output unless I have questions about how you computed your numbers.

Don’t make your appendix an integral part of the paper.  Put everything that you want me to see in the text or in your tables.

Tips for writing with numbers:

(a)    use the word ‘percent’ rather than the symbol ‘%’;

(b)    never begin a sentence with a numeral–spell out the number or use an introductory word;

(c)    don’t spell out numbers larger than ten;

(d)    don’t spell out any number with a decimal point or with the word ‘percent’;

(e)     don’t use more than one digit past the decimal point in any percentage.

Other tips for writing for me:

(a)    use parallel constructions (if two phrases or clauses are connected by ‘and’ or ‘or,’ they should have the same structure);

(b)    make sure your subject and verb match in number;

(c)    don’t say, “the employee and their supervisor”–make sure that pronouns match their antecedents;

(d)    remember that, for almost all uses in papers for me, ‘affect’ is a verb and ‘effect’ is a noun;

(e)     use the spelling checker on your word processor.