STAT4026/STAT3926
Statistical Consulting
Semester 1, 2022
Week 1 Report Writing

John Ormerod, Jennifer Chan, Alyce Russell,
and Jean Yang

Consulting in the past has covered many interesting topics from
social sciences, health sciences, agriculture, geology, transport,
education, etc.
It is hard to suggest one report structure that is suitable to all
type of analyses and aims:
• Model/method suggestion, eg. Gpower package, imputation
method, etc
• Provision of R template with simulated data, eg GLM, and
• Real data analysis.

Model suggestion:




GPower package support

The aim here is to provide some general guidelines and you may
not include all the suggested sections.
Note: Before writing your report, you may ask the clients for
data (if analyses are required), some written-down research
questions and research papers (if relevant).

Report title.
Under your report title, remember to include
• names of the lead, note taker, other members and the clients (if not in
the title), and
• date

Executive summary (1 page max).
– Short description of the problem(s).



– What are the main findings and key figures if appropriate?

– What is the practical (biological/clinical) relevance?
Some domain knowledge from the clients.
Background/problem.
– Longer description of the scientific problem(s).
Provide background, experimental designs and the objectives/aims.
– Translation of the scientific problem into a statistical problem.

– Data summaries and visualisation (may be a separate data section:
sample size, number of variables with description, data transformations)
Line plot of continuous X factors to show the trends (over time).

These plots serve to provide check and justification to include some
significant factors and their interaction for testing and/or modelling.

Other plot types for categorical (ordinal here) Y variable:

Data transformation
Eg. Feed Conversion Ratios (FCR)

Often provided from clients. Or you may transform the data to fulfil the
normality assumption but interpretation (from the client’s of view) may be
harder.
Data aggregation
Reduce data size but may smooth out noise or periodic effects, eg from
daily to weekly.

Missing data and imputation (if necessary), etc.


Note: provide title with important information and var labels for all tables
and plots.
Limitation of data


Analysis.
– What statistical tools, such as models and measures, are used and why.
Hypothese test?
Regression?
ANOVA? Repeated measure ANOVA to include time effect?
Logistic regression?
Ordinal logistic regression?
Parametric or non-parametric?
Univariate or multivariate?
Include random effects, nested effects or interaction effects?

How to write the model?
Using formula?


Or less technical using some vocal description?

– What are the results?
Present in tables?

With coefficients?
In figures or heat maps?

With p-values only?
In formulae?
To generate an aggregate engagement score:

– Interpret the results (to clients or statisticians?).

Interpret coefficients?
Aim for factor significance or/and the effect size?

Conclusion.
– Assumption check and shortcomings to the analysis.

– Summarise important founding and practical relevance.

Reference (if applicable).
Appendix.
Codes, more results, plots, technical details, model checks, time sheet, etc.


Consulting report marking rubric

riteria Fail Pass Credit Distinction High Distinction
Translate
between the
scientific
question and
the statistical
formulation
Wrong Slight misunderstanding
of the scientific question.
Simplistic, or formulated
to a limited capacity
Appropriate formulation. 1) Appropriate formulation.
2) Shows good
understanding of problem.
3) The specific sub-
formulation facilitate clarity
to the report.
Analysis Incorrect analysis Direct application of
techniques
No real justification Accurate and
appropriate
1) Accurate and
appropriate that is easy to
understand for the client.
2) Assumptions checked
and clearly communicated.
3) Robust analysis if
required.
Presentation Figures do not match
analysis. Errors in the
Figures. Figures do not
match data.
Not informative figures 1) Informative figures.
2) Axis labels, headings,
and legend.

1) Informative figures.
2) Axis labels, headings,
and legend.
3) Visually pleasing.

1) Clarify of the report -
headings and paragraphs
are used appropriately.
2) Informative figures with
clear labelling.
3) Report has limited
redundant information.
4) Visually pleasing.
Reproducibility Code does not match
figures/analysis.
Too many customized
elements for lecturer to
easily modify the code to
get it to run.
Readable, but not fully
reproducible.
1) Reproducible with
minimal changes.
2) Time sheet.
3) Organised document
structure
1) Fully reproducible with
commenting.
2) Clear recording of how
time was spent.
3) Organised and
documented file structure.

Discussion:
1. What is the difference between performing a hypothesis test and a
regression model? Which one should we suggest under different scenarios
if the clients have no idea?
2. If a client present a questionnaire with many items and ask how to
analyse the data, what will be your three questions to ask and why?