How to Write a Statistical Paper
In any statistical study, analyzing the data is only half the battle. You must then write a paper that outlines your research questions, describes the data collected and methodology employed, presents the results of the analysis and reports the conclusions and applicable recommendations. Most papers that use statistical analyses, whether written for a business or for an academic journal or conference, contain these sections. Following a logical sequence of steps will make preparing a statistical analysis report a more manageable task.
1). Introduction:
Introduce your paper by describing the subject at hand, stating the key questions posed by your research. If your paper is for an academic conference or journal, you also should describe how your paper will fit into the broader body of knowledge about the subject matter your paper will explore.
Summarize the existing research in your subject area by writing a literature review section. The literature review examines previous studies in a specific field and describes how the research in this field has developed.
Describe your statistical method. Explain the advantages of your methods. Briefly summarize your results.
Keep in mind most people only read introduction. So you need to include all essential points of the paper in the introduction.
2). Methods:
Describe the data collected and statistical analysis methods you employed. If your data came from a survey and you plan to analyze the data using “factor analysis”, describe the questionnaire and its purpose. Provide an explanation of “factor analysis” and why this method is most appropriate for your statistical study. This is usually the first part of the paper that one writes.
3). Results:
Report the results of your statistical analysis. Sections 2 and 3 will form the main body of your statistical paper. Use data tables and graphics, such as pie charts, bar charts and scatter plots to describe results. Pie charts are excellent for summarizing financial data, while bar charts effectively demonstrate differences in the data. The text of your paper that refers to a table or chart should briefly summarize the most important findings in the visual element, not simply regurgitate your data. In addition, the Online Writing Lab at Purdue University advises cautious use of statistical jargon. Briefly explain statistical procedures, if necessary.
4). Conclusion and Discussion
Complete a conclusions or recommendations section. Highlight the major findings of your statistical study, making sure that your conclusions and recommendations are supported by your analysis.
5). Appendix:
Tips for writing exploratory statistics:
Usually there is no good way to write a statistic. It rarely sounds good, and often interrupts the structure or flow of your writing. Oftentimes the best way to write descriptive statistics is to be direct. If you are citing several statistics about the same topic, it may be best to include them all in the same paragraph or section.
The mean of exam two is 77.7. The median is 75, and the mode is 79. Exam two had a standard deviation of 11.6.
Overall the company had another excellent year. We shipped 14.3 tons of fertilizer for the year, and averaged 1.7 tons of fertilizer during the summer months. This is an increase over last year, where we shipped only 13.1 tons of fertilizer, and averaged only 1.4 tons during the summer months. (Standard deviations were as followed: this summer .3 tons, last summer .4 tons).
Some fields prefer to put means and standard deviations in parentheses like this:
Group A (87.5) scored higher than group B (77.9) while both had similar standard deviations (8.3 and 7.9 respectively).
If you have lots of statistics to report, you should strongly consider presenting them in tables or some other visual form. You would then highlight statistics of interest in your text, but would not report all of the statistics. See the section on statistics and visuals for more details.
If you have a data set that you are using (such as all the scores from an exam) it would be unusual to include all of the scores in a paper or article. One of the reasons to use statistics is to condense large amounts of information into more manageable chunks; presenting your entire data set defeats this purpose.
At the bare minimum, if you are presenting statistics on a data set, it should include the mean and probably the standard deviation. This is the minimum information needed to get an idea of what the distribution of your data set might look like. How much additional information you include is entirely up to you. In general, don’t include information if it is irrelevant to your argument or purpose. If you include statistics that many of your readers would not understand, consider adding the statistics in a footnote or appendix that explains it in more detail.
Writing Statistics Plainly
In general, you should always ‘translate’ your statistics into some understandable form for your reader.
Poor example: “A t-test (t = 3.59) showed that the two groups were significantly different (p<0.01)."
The example above is complicated and hard to read. It's better to say something plainly first, then provide the statistical evidence afterwards:
Better example: Women scored higher than men on the aptitude test (t = 3.89, p < 0.01).
In the second example, your reader understands the relationship, it's not filled with jargon, but all of the same information is presented.
When using a complicated inferential procedure that your readers would be unfamiliar with, explain it. It may be necessary to go over it in detail. You may want to cite who used it first, and why they used it, and explain how it is applicable to your situation. A footnote or an appendix is a fine place for such an explanation.
If you include statistics that many of your readers would not understand, consider adding the statistics in a footnote or appendix if you can, especially if it is not central to your argument.
Writing Statistics Accurately
If you aren't sure how to calculate a particular statistic, either find out how, or don't use it. Along the same lines, never plug in numbers into a computer program, such as SPSS, and think that the output is "correct." Computer programs don't think for us; they simply allow for fast calculations. They cannot and do not interpret results. You should never interpret the results of a statistic that you don't fully understand. This is extremely important.
When in doubt, keep it simple. If the only thing you can say for certain is that the mean of one group is higher than the mean of another group, then that is fine. This is evidence, albeit it's not as strong as other types of evidence.
Remember that inferential statistics can never "prove" anything. You should think of statistics as a body of evidence (much like a fingerprint at a crime scene) that provides support for your argument. Sometimes it can be used as primary evidence or sometimes it is used in a more supporting role.
Focusing on Statistics
How you frame the use of your statistics is extremely important. In a more scientific field, you'll probably want your statistics as a focal point, but in other fields (say politics, for instance) you may use statistics to support a stance or policy, but it may be only one of many reasons for that policy. Knowing how your audience will react to statistics should affect how you use it. If your audience doesn't use a lot of statistics, you probably shouldn't make statistics the focal point of your argument, or if you do, you need to be very good about explaining the logic behind your statistics.
Acknowledgement
This article is modified from the article written by Shane Hall http://www.ehow.com/how_8245559_write-statistical-paper.html, and the articles “Writing With Inferential Statistics”, and “Writing With Statistics” published in Purdue Online Writing Lab.