Statistical Report Writing

1
Statistical Report Writing

• Martin Hazelton

2
Lecture Outline

• The Structure a Statistical Report
• Citations
• How to Start Writing
• English Usage
• Tables and Figures
• Mathematical and Statistical Notation
• Appendices
• Questions

3
The Importance of Structure

• The preparation of a scientific paper has almost
  nothing to do with literary skill. It is a
  question of organization.
• - Robert A. Day, How to Write and Publish a
  Scientific Paper

4
Structure of a Statistical Report

• Title
• Abstract
• Introduction
• Background Material (if necessary)
• Statistical Analysis
• Conclusions and Discussion
• References ( bibliography)
• Appendix (if necessary)

5
Title

• Aim describe contents of report clearly and
  concisely.
• Avoid using abbreviations and jargon in title.

6
Abstract (or Summary)

• Aim to summarize report in about 100 or 200
  words.
• Include main objectives, statistical methods
  used, summary of most important results, major
  conclusions.

7
Introduction

• What is the problem? What data are available?
  Describe it in a way that could be understood by
  any of the other 3S8 students.
• Why is it important?
• What work has been done on this type of problem
  before?
• What type of statistical approach are you going
  to take?

8
Background (if necessary)

• If the problem studied requires a time-consuming
  description, it may be better to give only a
  rough outline in the Introduction and to give
  details in this section.
• If the project involves a lot of new
  statistical methodology (i.e. material not known
  by other 3S8 students) then it can be covered in
  this section.

9
Statistical Analysis

• Describe statistical methods used (see next
  slide)
• Give results - fitted models, parameter
  estimates, confidence intervals, ANOVA tables
  etc. as appropriate.
• Say what statistical package was used, but do not
  spend time describing the particular operation
  that package. (E.g. Do not say, The data were
  loaded into R, then the lm() function was used.)

10
Describing Statistical Methods

• Assume reader has statistical knowledge of an
  average 3S8 student.
• Hence, if using standard statistical methods
  (e.g. multiple regression) there is no need to
  describe theory of the method just details of
  its application (e.g. fitted model, F and
  p-values as appropriate).
• New techniques will need to be described. You
  will need to show an understanding of what they
  do and how they work, but not technical detail.

11
Conclusions and Discussion

• Provide interpretation of statistical results.
• Ensure that your conclusions address the original
  aims of the analysis.
• How do your results fit into a broader context?
• Discuss possible extensions, generalizations etc.
  of your work.

12
References

• List of references cited in report.
• Journal article format
• Bloggs, J. and Smith, R. (1987). Variable
  selection in Cox proportional hazards models.
  Journal of the Invented Statistical Society, 34,
  11-30.
• Book format
• Jørgensen, B. (1993). The Theory of Linear
  Models. Chapman Hall New York.

13
Citations

• You can cite a text (book, journal article etc.)
  to support a statement, or provide the reader
  with a source of further information.
• Several styles for citations - simplest is to
  cite by author name(s) and date. E.g. Non-linear
  regression models have proved useful for the
  analysis of similar data sets in agriculture. See
  Watson (1987), for example.
• Use Bloggs et al. (1998) if 3 authors.

14
How to Start Writing

• Decide what points you want to get across in the
  section you are writing.
• Jot down a skeleton plan for the section,
  describing how you will move from one point to
  the next.
• Only then begin to write full paragraphs.
• Moral Dont jump in without planning first.

15
English Usage - A Nice Quotation

• One should not aim at being possible to
  understand but at being impossible to
  misunderstand.
• - Quintillian

16
English Usage - Rules of Thumb

• Best to avoid writing in the first person (i.e.
  using I) in statistical reports.
• Either write in the passive voice (e.g. The data
  were analyzed using logistic regression.) or
  using we (e.g. We analyzed the data using
  logistic regression.
• Try to avoid very long, convoluted sentences.
  Split long sentences in two where possible.

17
Correct English Usage

• Write in proper sentences - each sentence must
  contain a verb.
• Check spelling with a spell checker (but do proof
  read as well!)
• Data is a plural noun - the singular is datum
  
• Get the full stops in the right places when using
  abbreviations e.g. (not eg.), i.e. (not ie.), et
  al. (only a full stop after the l).

18
Tables and Figures

• Include in the main body of the text only those
  tables and figures that you think are essential.
• Each table and figure should be numbered (so you
  can refer to it in the text).
• Each table and figure should have a caption
  (describing the content of the table or figure).
• Keep tables as simple as possible.
• Comparison easier across table columns than rows.

19
Mathematical and Statistical Notation

• Keep notation consistent throughout report.
• Try and copy a good convention (from a book or
  from lecture notes, for example).
• Number any equation that you wish to refer to in
  the text.

20
Appendices

• An appendix is optional.
• If the data set is not too large then including a
  print out in the appendix can be useful.
• It can be used to include tables and figures of
  marginal importance.
• It can be used to include (annotated) computer
  code. (This should not appear in the main body of
  text.)

21
Examples What is wrong with?

• The regression of yield on the covariates crop
  density, temperature and humidity (which were
  stored in an S-Plus data frame my.data) was
  motivated by analysis of similar data sets by
  Peterson et. al. (1994).
• Age is not significant in model 2, and so we do
  not consider it any further.
• The p-value of 0.014 proves that age has quite a
  strong effect on the risk of catching the
  disease.

22
Checklist (i)

• Is report well structured? (Suitable numbered
  sectional headings employed?)
• Is report of an appropriate length? Should be
  about 20 pages i.e. about 5000 words with
  additional tables and figures as required.
• Have all cited works been included in the
  reference list?
• Has the report been spell-checked and proof read?

23
Checklist (ii)

• Is the notation consistent throughout?
• Are equation numbers, table numbers and figure
  numbers consistent?
• Does the title accurately characterize the
  report?
• Do the conclusions drawn at the end of the report
  address the aims of the project?