Homework answers / question archive / Final Paper Overview Your final paper is due May 15rd at 11:59:59 PM

Final Paper Overview Your final paper is due May 15rd at 11:59:59 PM

Statistics

Share With

---

Final Paper Overview

Your final paper is due May 15^rd at 11:59:59 PM.  You are allowed to work in groups of up to three people.  You can of course work alone.

I expect to see four outputs:

1. A pdf or word documents with your written work.  This includes graphs and charts.
2. Your R code.  Either in txt or an R output file.
3. A zip file containing pdfs of each of your references (if they are academic papers)
4. Your actual data.  Preferably in a csv file.

I am requiring a minimum of 7 pages per person, not counting graphs, charts and tables.

The data that you are using must not come from an R package.  

Your paper must include the following components.

1.  An introduction
2. A literature review
3. Exploratory data analysis 
4. Methodology
5. Results
6. Final conclusions

If your project required you to do significant work on acquiring and managing data, you should also include this as well.

Introduction

Your introduction should be fairly concise.  ½ to 1 pages is a very reasonable length.  It should state what it is that you are doing in your paper. 

What question/problem are you addressing?  I.e. What hypothesis are you testing? 

Why should anyone care? 

Are there important policy implications surrounding your topic? 

Are you testing an important theory?

What relevance does it hold to the world?

You absolutely should be sure to include what your contribution is to the topic.  Included in this, is identifying what data you are using and what your main results are. 

Literature Review

 

            What do the various papers that you have read on the subject say about your subject?  Are they all in agreement?  Is there significant disagreement amongst them?  These should largely come from peer-reviewed journals.

Your literature review serves a few purposes.  Firstly, it shows that you did the necessary research.  That you put in the time to learn about your topic instead of just writing whatever based on your feelings.  One way to think about it is, if the introductory section served to answer “why should the reader care about this topic”, the literature review helps demonstrate “why the reader should care about what you say about the topic”.

            It also serves as a reference point for the reader.  If he or she wants to know more about a given topic, this is an excellent source of information.  You will likely find yourself using other paper’s literature review to find new papers to read.  Finally,  it is also a way to motivate your research.  For example, “in the past people have looked at abc regarding this topic, but we are going to focus on xyz”.   

            Two places to start your literature review are Google Scholar and EconLit.  From there you can either trace papers forward or trace papers backwards.  By tracing backwards, we mean starting with a paper, then using that paper’s references and literature review  to find your next readings.   This will obviously give you older papers.  To trace forward, using Google Scholar or EconLit to see which papers have cited the paper that you are reading.  This obviously gives you more current papers. 

 

Exploratory Data Analysis

Here you should describe your data more thoroughly.  Where did it come from?  What period does it cover?  Is it a time series?  Panel data?  How large is your data set?  Is there anything missing from your data?  Are there any obvious variables that should have been included?  What are the relative strengths and weakness of your data set.

You should include summary statistics.  At a minimum you should identify the mean, median, min value, max value and standard deviation.  Higher statistics such as skewness and kurtosis may be included as well (Especially if you are working with financial data!).  These are all probably best placed in a table.  Histograms and box plots are very useful here.

Besides single variable statistics, it is certainly appropriate to include bivariate and multivariate statistics and graphics.  Correlation tables and scatterplots for example.  If you are working with data that include a number of categorical or ordinal data, conditional statistics would be highly appropriate.

You also want to be sure to identify any features of the data that you changed.  Are you only working with a subset of a larger dataset?   

Methodology

            These next two sections will likely take up the major body of your paper.  You start by clearly stating your model(s) in both English and mathematical form. 

            Where did your model(s) come from?  Did you come up with it yourself?  Or more likely, are you testing someone else’s model(s) or a modification of someone else’s model(s).  Secondly, how are you going to test your model?  OLS?  Logistic regression ect.

            You also should use this time to lay out all of the assumptions of your model.  

Results

Finally, are the actual results of your work.  This is likely going to be regression output and interpretation.  Remember, there is nothing wrong with finding insignificant results! 

Included in this section are your table of coefficients and your regression diagnostics.  Are the error terms distributed normally?  Do the variance of the errors terms appear to be constant or not?  Are there any outliers that are significantly affecting your model(s)?  Does it make sense to remove them?  And if so, how do your results vary.

AT NO POINT IN YOUR WRITEUP SHOULD THERE BE ACTUAL R CODE.

 

Conclusion

            What did you find?  Was it different from what previous researchers found?  If so, why do you think this is?  Additionally, you should discuss what additional avenues of research should be pursued.   Is there anything missing from your research?

            Remember, for each of these sections, you should let the various papers that you are reading be your guide.  They will undoubtedly vary from the strict structure that I am laying out for you, but you should be able to read a paper and see “oh this is how they present their model” and “this is how they explain their results”.

 

Presentation

Additionally, you are required to present your work in class.  10-15 minutes is likely sufficient.   Prepare any tables and charts in PowerPoint.  We will have this on the second to last day of class, 5/6.  And If we have to go over, we’ll use part of the last day of class as well.   

Plagiarism, Citations and References

Don’t do it.  Quotes should be denoted via quotation marks and either in text citation, footnotes or endnotes.  If a quote is sufficiently long enough, you can indent the text as well.  When you are restating another author’s words you need to cite their work as well.

            Personally, I prefer footnotes and citations to endnotes.  This allows the reader to identify the source without having to flip back and forth.  Endnotes are particularly annoying when reading in pdf form on an Ipad or computer.  But if you prefer them for your own reasons, you are welcome to use them.

Miscellaneous Notes on Formatting Your Paper

This includes unstructured thoughts and comments on writing your paper. 

AT NO POINT IN YOUR WRITEUP SHOULD THERE BE ACTUAL R CODE.

Do not, do not copy and paste output from R into your word document and hand it in.

Call:

lm(formula = Murder ~ I(Population/Area) + Illiteracy, data = statedata)

 

Residuals:

    Min      1Q  Median      3Q     Max

-4.8468 -2.1707 -0.2961  1.6636  6.5214

 

Coefficients:

                   Estimate Std. Error t value Pr(>|t|)   

(Intercept)          2.8617     0.8326   3.437  0.00124 **

I(Population/Area)  -3.1999     1.6689  -1.917  0.06128 . 

Illiteracy           4.2682     0.6051   7.054 6.78e-09 ***

---

Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

 

Residual standard error: 2.582 on 47 degrees of freedom

Multiple R-squared:  0.5309,    Adjusted R-squared:  0.5109

F-statistic: 26.59 on 2 and 47 DF,  p-value: 1.885e-08

 

As you are conducting your literature review take note of how the authors present their work.  Do they put their findings in organized and consistently formatted tables?  Or does it look as if they simply copied and pasted their computer code into a Word document?

Any mathematical models that you present should be written in math notation.  Familiarize yourselves with Excel’s equation editor.

yi=β0+β1x1,i+…+βkxk,i+ei

 

 

Graphs should be properly labeled.  The units of measurements should be clear.  The graphs of the left are best described as a “dumpster fire”.  The ones on the right are professional. 

 

 

 

Rounding

 

A good rule of thumb is to use two significant digits.  That is the first two digits.  When you just use the raw output from R, it a) suggests a level of precision that is unrealistic and b) makes it difficult for the reader to read.

 

Example, take 0 .0005642896 and report 0 .00056 instead.

Suggestions

I find it best to start a research project by assigning a folder in my computer.  Then within that folder subdirectories for:

1. Articles
2. Data
3. Graphical and table output
4. Rough drafts and misc notes

You can take this a step further and use either dropbox or google drive.  Both dropbox and googledrive have R packages that will let you automatically save data to either. 

I find it very helpful when putting together papers/reports to save and output and graphs and charts I want to save beforehand.  Then I can simply drop them into a Word document with ease.  Smaller charts can be saved as jpegs and larger, page sized, output can be saved as a pdf.