How to write in science Shah Ebrahim

Transcription

How to write in science Shah Ebrahim
How to write in science
Shah Ebrahim
Co-Editor, International Journal of
Epidemiology
Professor of Public Health
London School of Hygiene & Tropical
Medicine
&
South Asia Network for Chronic Disease
Above all: You must be clear.
Why?
To help the reader and avoid
misunderstandings!
To clarify your own thinking
Overview of session
•  General principles
•  Nuts and bolts
•  The sections of research paper
•  Technicalities of writing for
journals
General principles
Characteristics of scientific writing
•  Appropriateness
•  Balance
•  Persuasiveness
•  Brevity
•  Consistency
•  Precision
•  Sincerity
• Clarity
The four stages of composition
1: Think
• 
• 
• 
• 
• 
Define purpose and scope
Who are you writing for?
What readers need to know.
Allocate available time
Who else is involved?
Manage your time
Phases and resources
The four stages of composition
2: Plan
•  Draft an outline
•  Effective beginning
•  Put topics in order
•  End
•  Set the outline aside
•  Liaise with others
Battle plan
The four stages of composition.
3. Write
• 
Write so that you can easily revise
• 
Use outline
• 
Try to complete your first draft of the
paper, or of a section, in one sitting
Write as you like
ம"க$%: &பா)*+,& ம"க$% வ,த ஏ0
இ,2யா 4மான6, தைர இற":‌6;&, ஓ=
தள?2@ Aைல த=மாC ஓD
4ப?2E:Fளான&
Migration is a source of economic gain
but also of losses (stuff from previous
intro). Cite IJE paper.
Avoid distractions
The four stages of composition
4: Check
Read draft aloud
•  Main points emphasised?
•  Anything essential missing?
•  Meaning of each sentence clear and
correct?
•  Think of the reader: style, vocab
Revise
•  Ask two readers to check your corrected
draft
•  Revise again
Turn to Exercise 1. Which text is
easier to read aloud? A or B?
WORDS MATTER!
EXERCISE: PRINCIPLES OF
WORD CHOICE
•  Look at the underlined parts in the
following sentences
•  Could the be written better?
•  What is the principle of word choice that is
violated
•  If you are not sure, guess.
•  Just change the words not the sentence
structure
1.  The odds ratio was drastically reduced
when adjusted for confounders.
2.  Participants were asked to fast for
several hours before venapuncture.
3.  The cells were exposed to with
lipoprotein-deficient serum for 48 h
4.  With inhalation of salbutamol, lung
function improved.
Principle?
1.  The odds ratio was attenuated when
adjusted for confounders.
2.  Participants were asked to fast for eight
hours before venapuncture.
3.  The cells were incubated with
lipoprotein-deficient serum for 48 h
4.  After inhalation of salbutamol, lung
compliance decreased.
Principle? PRECISION
1.  Blood samples were drawn from 5
female and 3 male children at 1, 2, 3 and
4 h following the initiation of dialysis.
2.  Rapid dehydration was caused by
elevated body temperature.
3.  Meningococcal infection causes
erythematous, peticial lesions in severe
cases.
4.  Sampling was indicative of suboptimal
participation of younger men
Principle?
1.  Blood samples were drawn from 5
girls and 3 boys at 1, 2, 3 and 4 h
after starting dialysis.
2.  Rapid dehydration was caused by
fever.
3.  Meningococcal infection causes a
red blotchy rash in severe cases.
4.  Sampling was showed poor
participation of younger men
Simplicity
Principle?
1.  After 4 h of interviewing, we ended the
interview process.
2.  Socio-economic status and mortality was
examined and found to vary considerably.
3.  Obesity decreased in families with lower
household incomes but smoking increased
in poorer households. Both the decrease in
obesity and the increase in smoking were
greater when only men were included in the
analyses.
Principle?
1.  After 4 hours, we ended the interview.
2.  Socio-economic status and mortality
varied considerably.
3.  Obesity decreased in families with lower
household incomes but smoking
increased. Both trends were greater
among men.
Principle?
Necessity
1.  This study assessed the feasibility of
integrating quacks into governmental
malaria treatment services.
2.  While the male focus group chewed the
fat, the female focus group demonstrated
child-cleaning practices.
3.  Lymphoedema isn’t perceived as
sufficient reason to spend money on
consulting a doctor.
Principle?
1.  This study assessed the feasibility of
integrating all health practitioners into
governmental malaria treatment
services.
2.  While the male focus group discussed
the principles, the female focus group
demonstrated child-cleaning practices.
3.  Lymphoedema isn’t perceived as
sufficient reason to consult a doctor.
Principle?
Appropriate register
Principles of word choice
•  Precision
•  Simplicity
•  Necessity
•  Appropriate register
General vs specific
•  Avoid clichés and over used words: .
(e.g. without doubt, deliver care, address a
problem, day-to-day basis, gold standard)
•  Use more accurate and precise words
when you can
•  Avoid trendy euphemisms
resource-poor = poor
policy endpoints = results
Exercise Correct these phrases.
Why is this needed?
five different villages
a specific example
available evidence
observed values
Correct these phrases. Why is this
needed?
five different villages= five dissimilar villages, five similar
villages
a specific example = an unambiguous example, a
detailed example
available evidence = evidence on hand, obtainable
evidence
observed values =observed measurements, observed
standards
Improving precision = clearer meaning
Exercise: Make these phrases
concise.
•  are found to be in agreement
•  has the ability to
•  in no case did any of the controls develop
lesions
•  mortality in patients in Group 1 was higher
when compared to patients in Group 2
•  population growth increased by 100%
Make these phrases concise.
•  are found to be in agreement = agreed
•  has the ability to = able to
•  in no case did any of the controls develop
lesions = none of the controls
•  mortality in patients in Group 1 was higher
when compared to patients in Group 2 =
mortality was higher in group 1 than group
2
•  population growth increased by 100% =
the population doubled
Sentences
7
Make the message the grammatical
subject of the sentence
The message is that the index case came
from Mumbai.
•  It was discovered that the index case
originated in Mumbai.
•  The index case originated in Mumbai.
Put the action in the verb phrase
Measurement of the median distance to
the closest health post was done.
We measured the median distance to the
closest health post.
Mass administration of antibiotics occurred
twice in 2004.
Antibiotics were mass administered twice
in 2004.
Avoid putting the action as a NOUN
An increase in chloroquine resistance occurred.
Chloroquine resistance increased.
With parasitaemia of longer duration or severer
degree cognitive ability deteriorates
progressively.
What is parasitaemia doing?
When parasitaemia lasts longer or is more
severe, cognitive ability deteriorates
progressively.
Verbs are the lifeblood of
sentences. To weaken them or
to omit them saps the sentence
of its liveliness and makes it
difficult and dense to read.
Avoid noun clusters
range of patient data communication
requirements
can have six! meanings:
•  Meet requirements for communicating patients' data
•  Meet data-communication requirements of a wide range
of patients
•  Meet the requirements for communicating data from a
wide range of patients
•  Meet the requirements for communicating data to a wide
range of patients
•  Meet the requirements for communicating a wide range
of data about patients
•  Meet the requirements for communicating data about
wide range of patients
Write short sentences
Do not cram too many ideas in one
sentence.
Success of this method crucially depends on timely
delivery of the drug and on correct dosing, as
delayed or incorrect administration would lead to
development of resistance in the population, which
we saw in Kosovo, where migrant workers were
missed and children overdosed, resulting in a fresh
epidemic, comatose infants and loss of trust in the
health services, something the government cannot
afford and must avoid in the future.
Point of view
Author’s point of view
+ active voice
+ more lively and natural
- does not make the topic the subject of
the sentence
- may be inappropriate if fieldworkers or
technicians actually did the work
- gets obnoxious with overuse
Point of view of the experiment or study
- passive voice
- a bit dull
+ makes topic the subject
+ emphasises what is important
To keep writing interesting, alternate:
•  Use ‚ we for sentences that move the story
forward, i.e., in the study design
subsection but not for methods of
measurement.
Paragraphs
•  Mini-story
•  One paragraph for each aspect of the subject
(topic)
•  Use transitions to make clear the continuity
between sentences
•  General approach: Overview first, then details.
•  To identify the topic, use a key term, and repeat
it exactly.
Patterns of organization for
paragraphs
Most to least important
Announced order
Pro – con
Funnel (Introduction section)
Chronological order (Methods & Results
sections)
Problem – solution
Continuity
•  Repeat key terms.
•  Use transitions to indicate the train of
thought.
•  Keep a consistent order.
•  Keep a consistent point of view.
Keep a consistent order
To establish the chain of transmission, we had to
collect venous blood samples from the
suspected 1index case, from his 2immediate
family, from his 3recent contacts in the village
and from 4trade partners outside the village. The
sample from the 1index case could be obtained
in hospital; to collect the samples from his
2immediate family, we visited his compound.
Obtaining samples from his 3recent contacts
required that we spoke to the village children to
identify the contacts, and to then visit them at
home. To collect samples from 4 trade partners
outside the village, we had to interview the local
taxi companies ….
Writing a scientific paper
•  Interactive work
•  Titles and abstracts
Titles and abstracts
Hallmarks of a good title
•  Accurate: use same key terms in title as
in paper
•  Complete: include all necessary
information – e.g. study design
•  Unambiguous: avoid abbreviations
•  Concise: omit unnecessary words
•  Important word first
•  Engaging (e.g., a question)
What is wrong with this title?
A brief account of smoking by Tibetan
migrants and modification by the effects of
living conditions, SEP and length of time
spent in India.
A possible improvement?
A brief account of smoking by Tibetan
migrants, COPD and modification by the
effects of living conditions, SEP and length
of time spent in India.
Smoking in Tibetan migrants to India is
associated with chronic obstructive
airways disease: a cross-sectional survey
The Abstract
EXERCISE 1
Abstract
Abstract
Abstracts
What?
Why?
How? Who? Where? When?
Findings?
Conclusion?
Only essential information following the
overall structure of the paper.
Structured vs. unstructured abstract
•  Structure focuses the mind
Facilitates conciseness
Is easier to read and retain
•  Write a structured abstract before you start
and again when the paper is finished
The Introduction
EXERCISE 2
Page 1
EXERCISE 1
3
1
2
4
Role in the Story Line
•  First step
•  Research question (hypothesis-testing
paper)
•  Message (descriptive paper)
•  New or improved method, apparatus or
material (methods paper)
•  Working metaphor: Funnel
Content
• 
• 
• 
• 
• 
• 
Known
Unknown
Question
Material or population
Study design
References: only the first, the most
important, the most recent
•  Newness, importance (optional)
No answers, results, implications
Funnel principle
S t a r t
b r o a d l y
N a r r o w s t e p by s t e p
to a focal
point
From the Known
to the Unknown
to the Question
Step 1: What is known
•  In first sentence, state general topic of the
paper
•  Explain background/motivation in a few
sentences or paragraphs
•  Within this framework, narrow down to
what is not known in the research area
Step 2: What is unknown
•  Brief statement, usually one sentence.
•  Important because it
indicates that work is new
links Known and Question, creating
the story line
Writing about what is unknown
Clearest if flagged:
… is unknown
… needs to be determined
… is unclear
... However,
it is not yet known how low social
support leads to increased mortality.... To
answer this question, we randomly allocated
participants to a social support intervention
or to usual care ...
Step 3: Your Research Question
•  Specific topic of the paper
•  Must inevitably follow from Known and
Unknown
Importance
Best placed at the begining of the
Introduction.
Make sure the importance of the work is
evident – state if necessary.
Remember modesty.
To develop the story line:
•  Start a new paragraph for each of Known,
Unknown, Question if the Introduction is
long
•  Use transition phrases and words
•  Repeat key terms
Length of the introduction
Typically 1 double-spaced A4 page, i.e.,
250-300 words
Do not exceed 2 pages or 600 words
Materials and Methods
EXERCISE 3
Materials and methods
• 
• 
• 
• 
• 
• 
What is the study design?
What is the independent variable/ intervention?
What are the dependent variables?
How were the data collected and analysed?
What is the sample size?
Could you redo this study from the methods as
reported here?
•  Readability?
Page 2
EXERCISE 2
Page 2
EXERCISE 2
Functions
•  Sufficient detail to permit evaluation and
replication of your work
•  Hypothesis-testing papers: what you did to
answer the research question
•  Descriptive studies: what you did to obtain
the message
•  Methods papers: (1) new method in detail;
(2) what experiments done to test the new
method
Role in the Story Line
The plot thickens.
Overview of the experiments or study, the
plan
Participants
•  Age, Sex, Race
•  Selection process
•  State of health or disease
•  Specific medical or surgical management
•  Ethical approval
What constitutes sufficient detail?
Study design (overview of the methods)
•  Questions asked
•  Independent variables = interventions
•  Dependent variables = measured
variables
•  Comparison groups
•  Sample size – prior estimates of
requirements made?
•  Methods, laboratory assays and
equipment
– Well-known: reference only
– Less well-known or modified: essential
features & reference
•  If in doubt, give more detail
Statistical analysis
•  Tests used and which measurements they
compare. Only reference uncommon tests
•  Program used including version/release
number
•  Sample size if it is not obvious from study
design
•  P values + 95% Confidence Intervals
Overall organization of M&M
Generic subheadings
Study subjects
Study design
Interventions/exposures
Methods of measurement
Calculations
Analysis of data
Ethical approval
The Results
EXERCISE 4
Page 3
EXERCISE 3
Page 3
EXERCISE 3
No strong evidence of associations between any
exposure and phlegm were found.
Functions
•  State results of experiments described in
M&M section
•  Direct the reader to tables and figures
•  In terms of story line: the plot unfolds
•  Working metaphor: snapshot
Content of the Results section
•  Report only results pertinent to the
research question stated in the
Introduction
•  Report results whether or not they support
your hypotheses
•  Present data in figures and tables and
refer to these; mention only the most
important data in the text
Results vs. Data
Data are facts, often numbers, obtained
from experiments and observations.
Raw data = e.g., all bodyweights
measured in a newborn population
Summarized data = e.g., mean plus
standard deviation
Transformed data = e.g., percent of
control
Results are general statements that
interpret data, i.e., the meaning of the
data.
Data can rarely stand alone. The result
should be stated.
Example: Data but no result
In the 20 control subjects, the mean resting blood pressure was 85 ±
5 (SD) mmHg. In the 30 tennis players, the mean resting blood
pressure was 94 ± 3 mmHg.
Revision A: Result but no indication of magnitude
The mean resting blood pressure was higher in the 30 tennis players
than in the 20 control subjects [94 ±3 (SD) vs. 85 ± 5 mmHg, P <
0.02].
Revision B: Result and general idea of the magnitude
The mean resting blood pressure was 10% higher in the 30 tennis
players than in the 20 control subjects [94 ±3 (SD) vs. 85 ± 5 mmHg,
P < 0.02].
Data
•  must be accurate.
•  must be internally consistent
For example, each percentage should show the
same level of exactness, such as 1 or 2 digits
after the period.
If a value is given in several places, i.e. in the
Results, in the Discussion, in a figure or table, it
must be same everywhere.
Statistical analysis results
•  Interpretation of confidence intervals
should focus on the implications (clinical
importance) of the range of values in the
interval
•  Avoid the term „statistically significant“ the strength of evidence for or against the
null hypothesis is indexed by the P value.
The smaller the P value, the stronger is
the evidence
Statistical analysis results
•  take a very sceptical view of subgroup
analyses in clinical trials and observational
studies.
•  In observational studies considerations of
confounding and bias are as important as
statistical significance
See: Sterne J, Davey Smith G. BMJ 2001;322:226–31
The Discussion
EXERCISE 5
Page 4
EXERCISE 4
Page 4
EXERCISE 4
Page 4
EXERCISE 4 (CONT)
Functions and Content
• 
Answer research question(s)
• 
Explain answers
• 
Contextualise
• 
Draw conclusions
Story Line: Denouement
•  What we have learnt and its implications
•  Hypothesis-testing paper: true / false
•  Descriptive paper: key features of the
observations described + implications
•  Methods paper: advantages and
disadvantages, applications
Question-answer match
Research question in Introduction:
Does mass administration of artemisinin
induce resistance in P. falciparum?
Answer in Discussion: This study shows
that mass administration of artemisinin
induces/does not induce resistance in P.
falciparum.
Giving credit to yourself and others
• 
Mention other people’s supporting results
These results thus lend further credence to X’s earlier suggestion
that …
Y previously hypothesised that … . In this study we found that a
similar situation emerges for …
• 
If your work pulls together a lot of loose
threads, say so.
This reconstruction, which incorporates and confirms the separate
observations of X, Y and Z (refs), includes several new
observations that provide a more complete understanding of …
• 
Be matter-of-fact.
Defending your answer
Necessary if
•  other answers have been proposed for the
question you asked
•  other answers are easy to imagine
Explain both
•  why your answer is satisfactory and
•  why others are not
Examples for defending an answer
“In this study we found that the effects of migration on
obesity occurred within 5 years. We believe that this
duration is credible as increases in dietary intake and
adoption of sedentary habits are rapid after migration.
However, much longer durations have been reported by
other investigators 6, 7. But their study populations were
subject to seasonal migration and substantial losses to
follow up. We suspect that valid estimates of migration
effects can only be established in relatively stable
populations, such as ours … “
“Apparent discrepancies between our human growth
hormone values and those of earlier studies27,48 may be
due to differences in the study design …”
Conflicting results
• 
• 
mention
explain
“This finding cannot be attributed to increased calorie
intake of urban men, since values were no different from
those of rural dwelling men. Rather, it appears that
physical activity levels were higher in rural men, allowing
them to maintain their body mass index rather than
become obese.”
Newness of your findings
•  should be established in the Introduction
•  To remind the reader, contrast:
“Associations with social isolation and self-ratings
measures of ill-health have been reported 38-40. In this
study, we report a comprehensive, clinically validated,
assessment of mental health status...”
Unexpected findings
•  signal at the beginning
•  explain
…A surprising finding was that bug-o-kill did
not eliminate the parasite, but that it induced
somnolence in human subjects. We suggest two
possible explanations: … If these explanations
are correct, they imply that bug-o-kill may be
used to treat sleeplessness …”
Shortcomings and presuppositions
Explain
•  limitations
•  weak study design
•  basic assumptions
If the explanations are only 1-2 sentences long,
think harder about why the work could be
considered flawed or inadequate. Reviewers will
come up with a very long list for sure!
Importance of your work
• 
• 
• 
• 
Obvious importance may not need to be established.
Applications
applications (X will/can … do Y),
Implications
our results suggest/ imply that X may/might
Recommendations
we recommend that … should/must
Speculation
gives rise to the speculation that X does Y
Place at the end of Discussion section.
Organization of the Discussion
•  Start: answer research question; support
answer with results
•  Continue: (1) put your work in context, (2)
discussing limitations and assumptions.
•  Finish: Indicate applications/implications.
How not to begin the Discussion
•  with a second Introduction
•  with a summary of the results
•  with secondary information
How to end the Discussion:
Make a point
•  Signal the end (“In summary, …” “In
conclusion, we …”)
•  Indicate the importance of the work
by stating its uses.
How not to end a Discussion
•  stating that further research/studies are
needed
•  claiming research territory by staking out
your next project
•  claiming that your study is the only/first
one to answer the research question
Subsections
Subheadings are not necessary, unless:
•  Discussion exceeds 1000 words
•  Discussion contains 3+ subsections each
dealing with a separate major topic
•  But can be useful in drafting to ensure you
don’t overlook something
Conclusions
•  Link conclusion to goal of your study
–  e.g. should provide the answer to the question
posed in introduction
–  should make it clear whether hypothesis was
or was not supported by the data
•  Should not over-reach into areas not
covered by the research
•  Beware data  conclusion mismatches
Acknowledgements
•  Contributors
•  Funding sources
•  Reviewers
A brief comment on plagiarism
•  Don’t do it – editors check!
•  Word for word cut and paste is plagiarism
•  Using quote marks “There is no more
important question than finding the best
means of controlling the obesity epidemic
in China” Popkin 2011.
is fine.
Technicalities of writing for
journals
Step 1: Writing up – formal aspects
One common, reader-friendly font
No more than 2 sublevels
Ask colleague to proofread
Step 2: Fitting your manuscript to
the formal requirements of your
target journal
•  Read the instructions.
•  Use a reference software (e.g. Endnote)
for ease of changing citation format
•  Stick to word count limits
•  Be realistic about where to send your
manuscript
•  Check and recheck references.
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi
•  Get permission to reproduce text or
graphics from a previous publication
Step 3: Get all authors on board
BEFORE you submit
•  Authorship = must fit the journal’s criteria – being
the boss is insufficient
•  Agree on the final text and send it to each
author.
•  Decide on the order in which authors are to be
named.
•  Decide who gets acknowledged.
•  Agree on corresponding author.
•  Describe each author’s role, collect signatures
•  Declare conflicts of interests, if any.
•  Make sure you have copies of ethical approval
documents.
Step 4: Check the STROBE* guidelines
http://www.strobe-statement.org/
•  These guidelines are intended to help
authors to produce good descriptions of
epidemiological studies
•  You should certainly ensure you cover all
the points
•  Several journals have decided to use
these guidelines in reviewing
epidemiological papers
*STrengthening the Reporting of OBservational studies in Epidemiology
Step 4: Submit paper for peer review
•  Write a short!!! cover letter
•  Include authorship statement, all authors’
contact details, and, as required,
declaration of interests, the copyright
transfer and suggestions for reviewers if
applicable.
•  Expect acknowledgement of receipt within
a week.
•  Expect feedback after 8-12 weeks.
Step 5: Revision
•  Stick to the time frame given by the editor.
•  Make your revision as easy to evaluate as
possible.
•  Include a revision with changes tracked and
one without.
•  You do not always have to agree with the
reviewers, but you must comment on each
suggestion!
•  Expect final decision in another 6-8 weeks.
Step 6: Acceptance
•  You should, again, ensure that all authors
have the version of the paper that is to be
published.
•  Some journals will inform you when the
paper is scheduled to appear.
•  Make sure the editors have contact details
of at least two authors!
Step 7: Proofs
The responsibility for correctness lies with
you, the author!
Return the corrected proofs as soon as
possible.
Most common step –
REJECTION!
•  No-one likes a rejection letter
•  At IJE we reject 90+% of manuscripts
•  Everyone gets upset – so don’t take it
personally
•  Leave it a week and then re-read your
paper and the reviewers’ comments
•  Discuss with colleagues
•  Try again – taking the comments into
account – it may go the same reviewers!
Summary
• 
• 
• 
• 
• 
• 
• 
1: Writing up – formal aspects
2: Fitting manuscript to journal
3: Get all authors on board
4: Submit paper for peer review
5: Revision
6: Acceptance
7: Proofs
GOOD LUCK!!