Write a 14-15 page research article based on either the PSID given to you in class or the National Educational Longitudinal Survey (NELS).   The article should examine an important topic and hopefully one that has not been previously researched (though, for this class article, this need not be the case).  Choose at least 12-15 independent variables for the article and try to include more than a single dependent variable for what you’re examining. For example, you may be interested in issues related to work or how financially sound individuals are given their family backgrounds.  You may look at a number dependent variables including hours of work, whether or not the individual is living below the poverty line, and income or family income relative to the poverty line (a continuous variable that takes into account family size).  You may choose to measure the adolescent effects of welfare receipt, hours that the parents worked, and types of occupation of the head of household.

I would like you to use some of the different skills you’ve learned in the class to examine your topic.  For example, I would like you to run interactions between independent variables, run quadratic or cube functions, use log dependent variables, use Chow tests or likelihood ratio tests to determine if groups are different or adding variables adds anything to your model.  Run regressions with by adding sets of variables to see how your main independent variable changes when adding control variables.  You will thus have several columns of results as you add more and more variables.  The adding of variables should make some logical sense – don’t simply add them and give no reason for why you’re adding them. 

The following is a guide to how to write the article.

1. Statement of the problem.  What are you studying and why is it important?

2. Literature review.  What have others found to be important in the study of your problem.  The problem you are examining is your dependent variable.  For example, have others found adolescent poverty to be an important indicator of adult poverty.  If they have then list the authors that have found poverty to be important.  Perhaps in other studies, poverty was not found to be important.  List the authors that have not found poverty to be important.

Do not make an annotated bibliography in this section.  That is, do not list the articles and give descriptions of each of the articles.  Categorize the articles by subject area, such as poverty, and briefly discuss the findings of these studies.  If several of the studies have the same findings, discuss them together (within the same sentence). 

Remember that you are trying to explain your dependent variable as best you can.  You therefore do not want to only focus on the one or a couple of independent variables that you believe are most important in your study.  You should also focus on those variables that others have found to be important in researching your dependent variable, and include what they have found for these other independent/control variables.  

Limit the number of web sources you use to a few.  I would like you to examine the journals for what others have found regarding your dependent variable.  Also, limit citing sources that come from other sources. This means, do not say, "Johnson (1999) as cited in Vartanian (2006)".  Go to the original if you want to cite it. 

Have at least 10-12 sources for this article.

3. Theory.  Discuss the theoretical foundations for what you’re examining.  For example, if you’re studying persistent poverty, you may wish to examine culture of poverty theories or theories related to lack of economic opportunities.  I often examine alternative theories to determine which theory is best supported by the data and statistical analyses.  Explain where the theories come from?

4. Data sources and hypotheses.  Explain the data you'll be using for your study.  Discuss where the data comes from and the types of variables included in the data (in a general way).  Include the number of observations, perhaps the number of observations in the control and experimental groups (if there are such groups), the year or years the data were collected and the population from which the sample was taken from. For what populations can you generalize your results?

Hypotheses.  Given your problem statement, your literature review and your theories, what hypotheses follow?  Relate your hypotheses, your dependent and independent variables, to previous literature and theory.  Why are you hypothesizing what you're hypothesizing?  Perhaps you're hypothesizing the positive or negative relationship between variables because no one else has examined these relationships and these relationships are important for some theoretical reason.  You should come up with a logical reason why the relationships you're examining are important.  Or perhaps you're reexamining the relationship between variables -- others have examined this relationship but you have a different data source and will see if this relationship holds up for this different data.  Remember that an unbiased b coefficients doesn’t mean that you don’t need to replicate studies.  To find the population B coefficient, replication is necessary. You would like to verify that what appears to be going on in one sample is also going on in your sample -- you will replicate to verify (I believe that Ronald Reagan used to say something like this
about the Soviet Union).

Explain the variables you’ll be controlling in your statistical models and the importance of controlling these variables.  Cite other studies that have also used these variables as control variables.

5. Describe the types of statistical tests you’ll be using to examine the relationship between your independent and dependent variables.  In other words, you’ll indicate that you’re using a logistic regression, ordinary least squares, or other type of statistical models.

6. Results.  In this section, first report on the descriptive statistics for the study.  What are the means and standard deviations for the dependent, independent and control variables?  Create a table with this information and describe what you have found in the text of the article.  Next, report on the outcomes of your statistical tests -- either the OLS multivariate regression or the logistic regression, or whatever other type of statistical analysis you're using.  Was there a relationship between the independent and dependent variables or not?  Was the relationship in the correct direction?  Did your tests give support to your hypotheses?  Or must you accept the null hypothesis (or, in other words, you cannot give support to your hypotheses)?  Was the effect small in spite of being significant?  Exactly what is the relationship between the independent and dependent variables?  You’ll be able to determine this relationship by examining the b coefficients from the regression results. Next, I would you to present a table of predicted outcomes -- either probabilities if you are using logistic regression analysis or other types of predictions, using OLS or some other statistical analysis.  Talk to me about the type of predictions you should use for your paper and the SAS programming you will need to get these predictions.  If you are using indexed dependent variables, you will need to talk to me about getting predictions for these types of variables.  Generally, you will also need to choose an non-indexed variable also for your predicted outcomes.  Again, you will present at least three tables in the results section – one for the means and standard deviations,a second for the regression results, and a third for the predicted results.  Many of you will be presenting more than one regression table if you are also examining fully interactive models, or examining the effects of your independent variables on different groups.  I encourage you to be as creative as you can be in examining your models. 

In your tables, use no more than 3 decimal places for b, SE, and probability estimates.  Try to limit the number of tables to 5 or 6.  In your regression models, include all variables you've used in the regression so that I can more easily see what you're doing.  Use astericks to indicate statistical significance and put a note at the bottom of the table that indicates what the astericks mean.  Label each table explicity so that I know what kind of  regression you're using.  In the label, indicate the dependent variable.  Give each table a number (Table 1, Table 2).  You must write about each table that you present and include the tables you are discussing in the order that you discuss them.  Thus, don't discuss Table 3 first.  In your descriptive statistics, include all variables you've used in your regressions, including all independnet variables (included excluded categories for dummy variables) and dependent variables.  If you are doing analyses for separate groups (such as gender or race groups), include columns for means and standard deviations for these groups for each of your independent and dependent variables. 

7. Conclusion.  Go over what you were looking for in your analyses and what you found.  Do the two match up?  What is the significance (not statistical) of your findings?

Remember that you need to have transition statements between each of these different sections. Relate one section to the next, as you would in any paper.  Note that you may wish to order your paper somewhat differently than the way I’ve described it above.  Look for examples of how to write a research paper from the articles you’re reading this semester.  Follow their lead. Examine how they put together their literature review, how they report their results, etc, and their ordering of topic areas.

What to hand in:

A. The paper, with a title on a title page, and an abstract.  The paper should have 1" margins and use 12 point, Times New Roman font.  Papers over
    20 pages long will not be accepted.  Include a bibiography but these pages will not count as part of the 20 page maximum.
B. The SAS/SPSS/STATA output
C.  All programming that you've done to get your output.