Transcript Scientific Writing

Scientific Writing
Table:
Part-3
Summary of last session
• Title:
• Informativness without text: group Vs given name…
•
•
•
•
•
•
•
•
Topic or point
Independent variable(s) (X),
Dependent variable(s) (Y),
Animal or population, the material described, or both (Z).
Category term
Headings
Dependent and independent variables
Use the same key terms in the title, the column headings, and
the text of the paper
• Informative
Headings
• Do not omit the column heading
that states the name of the
dependent variable [for example,
“Recovery (%)” in Table 5],
even in simple tables that have
only one dependent variable that
is named in the title.
• It is clearest for the reader if the
dependent variable is named
both in the title and in the
column headings.
Subheadings
• In the column heading
the dependent variable and the unit of
measurement are in the main heading and two
sites in which this variable was measured are in
the subheadings. (See also Tables 2, 4, and 5.)
• Note that terms in column headings and
subheadings are singular, not plural (for
example, “Recovery,” not “Recoveries”).
lf necessary, number columns or rows to
help integrate the text and the table.
• A table can sometimes require extensive
explanations in the text, and extensive
explanations in the text sometimes require
frequent reference to a table.
• In such cases, it can be helpful to number the
column or row heads to help readers find their
place in the table faster .
lf necessary, number columns or rows to
help integrate the text and the table.
Building
A table
Title
Column
Row
headings headings
Fields
Foot
notes
• The purpose of a table determines how you construct
it: Form follows function. If the table's purpose is to
compare subjects at different times, for example, then
you will want to format the table so that this is easy to
do, by placing the times side by side in columns (Table
6.3).
Table Formats
• The rules on which words in a table title should
be capitalized will vary from journal to journal.
• Follow journal guidelines :
– roman or arabic numbers;
– centered or flush left table number, title, column,
headings, and data;
– capital letters and italics;
Building
A table
Title
Column
Row
headings headings
Fields
Foot
notes
BUITDING A TABTE:
The Field
If an intersection does not have a datum, the absence should be indicated by
three ellipsis dots (. . .) or an abbreviated notation explained in a footnote,
such as ND for "Not done" or NA for "Not available', or .,Not applicable".
The Field
• To save space, editors often allow some
abbreviations to be used in tables that they would
not allow in text.
• All abbreviations in tables (but nor standard
symbols for units of measurement) must be
explained in footnotes, even if they are
abbreviations explained in the text.
The Field
• Placement of standard deviations can be
difficult, especially in the case of several
columns.
– If placed to the right of the mean, reading and
comparison of standard deviations across rows are
hindered.
– Likewise, if the standard deviations are placed
below the mean, reading and comparison along
rows are hindered.
Tables: baseline, descriptive data
Women’s health Initiative, JAMA, 2002
n
Standard
deviation
Building
A table
Title
Column
Row
headings headings
Fields
Foot
notes
Footnotes
•
The footnote
1.
2.
3.
4.
5.
6.
defines the kind of data,
the number of patients studied,
The observation points,
the abbreviations,
the statistical significance level, and
the statistics used.
Footnotes
• For example, "12.1 +_ 0.3" could indicate a
mean of l2.l with a standard error of the mean of
0.3 or a standard deviation of 0.3 for the
distribution of the values represented by the
mean.
• Keep footnotes to a minimum. They should only
be used to explain essential details and
abbreviations.
The footnote defines the kind of data,
1. the number of patients studied,
2. the observation points
3. the abbreviations
4. the statistical significance level
5. the statistics used.
Footnotes
From: nurses health study
general accepted standards
• Although certain aspects of table format differ
between journals, some generally accepted
standards exist.
• Three horizontal lines are usually used to separate
parts of the table:
– one above the column headings,
– one below the column headings, and
– one below the data (to separate the body of the table
from the footnotes).
Tables: baseline, descriptive data
Three
horizontal
lines
Table 1. Base-Line Characteristics of the Women Who Underwent Radical Mastectomy and
Those Who Underwent Breast-Conserving Therapy.
Veronesi et al. Twenty-Year Follow-up of a Randomized Study Comparing Breast-Conserving Surgery with Radical Mastectomy for Early Breast Cancer
NEJM 347:1227-1232; October 17, 2002
Exercise : Checklist
•
•
•
•
Title
Headings
Field: comparison, precision…
Footnote ( 4 parts)
Exercise BMJ step by step
Table 4.1 is an example of a poor table.
• The title does not explain the
initiating stimulus to the observed
responses.
• It lists individual haemodynamic
variables rather than using a
category term.
• The "condition" is poorly
defined.
• All of the vertical and most of the
horizontal grid lines are
superfluous. The columns have
no indication of the units used.
• The results for cardiac output
show more decimal places than
the precision of the measurement
justifies.
Table 4.1 is an example of a poor table.
• The ± value is not defined (is it
standard deviation or standard
error of the mean?).
• No mention is made of the
number of participants studied.
• The changes run from top to
bottom rather than from left to
right across the page.
• Abbreviations are not
explained.
• No indication is given of any
statistically significant changes.
Consider now a revised version of the
same table (Table 4.2).
• The title describes the initiating
stimuli ("Induction of
anaesthesia and sternotomy")
and uses a category term
("cardiovascular").
• The superfluous grid lines are
eliminated. Changes run across
the table from the left-most
column.
• The subheadings ("Before",
"After", "During", and "After")
allow clear chronological
allocation of observation points.
Consider now a revised version of the
same table (Table 4.2).
• Units of measurement are
provided. The asterisks in two
columns are a clear signal of a
statistically significant difference
(the absence is a clear signal for
the lack of such a difference).
• The footnote defines the kind of
data, the number of patients
studied, the observation points, the
abbreviations, the statistical
significance level, and the statistics
used.
• This table can now stand on its
own. Your reader will be able to
obtain all the information they
need without having to refer back
to the text.
Next sessions
• Part 4:
– How Should Be Compared? (How to
complement results with tables ?)
• Part 5:
– What Should Be Compared?
Scientific Writing
how to complement
results with tables ?
Part-4
Akbar soltani, MD
Shariati Hospital
www.soltaniebm.com
www.ebm.ir
www.avincennact.org
GETTING STARTED
• Many beginning investigators make the mistake
of writing their results section around their
tables and figures.
• The resulting manuscript reads like a boring
tour of an art museum.
• Picture 1 is by van Gogh. That sculpture is by
Matisse. The next painting was painted in 1904.
• Table 1 shows. . . . Figure 1 demonstrates. . . .
The multivariate results can be seen in Table 2. .
. . Figure 2 is the survival curve.
GETTING STARTED
• While there is nothing intrinsically awful about this
type of paper, it violates a key rule, that the text, tables,
and figures should complement one another.
• What does complement mean?
• If the table gives the participants' mean weights and
cholesterol levels, and the proportions who had
diabetes and hypertension, then you might say this in
the text:
• Participants were generally obese with a high
prevalence of cardiovascular risk factors (Table 5.1).
What does complement mean?
Complement
means
Logical
structure
TABLES OF LISTS
• Some tables are just lists of items,
such as diagnostic criteria or types of
bacteria that cause urinary tract
infections.
• The essential decisions are what to
include and in what order.
TABLES OF LISTS
• A list may not have an intrinsic order, such as
the causes of hyperkalemia, or it may have an
order that is essential to the point you are trying
to make, such as the frequency of symptoms in
a patient with Cushing's syndrome.
• Beware of imposing an order that confuses your
point. It makes little sense, for example, to
alphabetize urinary tract pathogens (Table 6.8).
TABLES OF LISTS
• The sequence in such tables should have a
readily grasped logic:
–
–
–
–
descending order of frequency;
grouping by body systems;
chronologic order,
or some other clear basis.
•
•
A better approach is to list the pathogens by frequency, perhaps subcategorized by
type. A more informative title and headings also help (Table 6.9).
Table 6.9 demonstrates an important rule: Avoid using percents of percents. Each
percentage in a column should refer to the same denominator. If you want the reader
to notice that half of the gram-positive species were Staphylococcus aureus, then say
so in the text: "We cultured several different organisms (Table 6.9). Staphylococcus
aureus accounted for half of the gram-positive culture results."
•
Note that the percentages for the two most common types of gram-negative pathogens
do not sum to the total for all gram-negative bacteria; this immediately informs the
reader that other species were seen. If these species matter, they could be included in
a footnote. The percentages for specific gram-negative (and gram- positive)
organisms are offset slightly, to indicate that they are a part of the total for that group.
The numbers of each pathogen have been omitted intentionally; they would be
included only if you wanted readers to be able to check your arithmetic.
TABLES OF SUBJECT
CHARACTERISTICS
• Ask yourself what readers will want to know
about the subjects. How does the example
(Table 6.10) look?
• At first glance, the table may look fine. But it is loaded
with problems.
• The title is generic: which subjects and how many?
Stating the proportions of subjects who were male, and
then doing the same for female subjects, is not
necessary.
• The extraneous variables and unnecessary
precision are distracting; will anyone really care
about the subjects' mean shoe size, or that
exactly 8.63% of subjects had a history of
diabetes?
• The table mixes dichotomous data, such as
history of diabetes, with continuous data such as
body weight.
• There is an undefined abbreviation (CHD), the
columns are not labeled, the units are not provided for
age and body weight and they do not make sense for
caloric intake, and the meaning of the plus-minus
values is not specified.
• Exercise: Consider the improved version, Table 6.11.
What does complement mean?
Complement
means
Logical
structure
Overall
Compare
result
Contrast
(similarities)
The most
(differences)
Common…
• The text that accompanies the table might simply read,
"There were similar numbers of men and women in the
study (Table 6.11); 33% of the subjects were over 65
years of age, and 25% were more than 10 kg above
ideal body weight. Most were free of chronic medical
problems."
• An acceptable alternative, especially if the exact
numbers do not matter very much, is to list the
percentages and means (Table 6.12).
What does complement mean?
Complement
means
Logical
structure
Overall
Compare
result
Contrast
(similarities)
The most
(differences)
Common…
Stratify
• Occasionally, it is worthwhile to divide
(stratify) the information on the subjects into
two groups (e.g., men and women). This type of
table is most useful if there are important
differences between the two groups; which
should be pointed out in the accompanying text:
• Text: "Diabetes was 40% more common
among women, whereas male subjects were
more than twice as likely to have reported a
history of heart disease."
What does complement mean?
Complement
means
Logical
structure
Overall
Compare
result
Contrast
(similarities)
The most
(differences)
Common…
Emphasize
• The text can also be used to emphasize other
aspects of the sample, such as the fact that
the numbers of men and women were
similar, or that the overall mean age was 64
years (Table 6.13).
What does complement mean?
Complement
means
Logical
structure
Overall
Compare
result
Contrast
(similarities)
The most
(differences)
Common…
Emphasize
categorize
outcomes
TABLES THAT,TELL WHAT
HAPPENED
• If the manuscript reports the results of a
randomized trial, then the characteristics of the
subjects need to be presented for each of the
study groups.
• Percentages-rather than numbers-of subjects
may be easier to follow.
• This is especially true if the numbers of subjects in the groups vary
substantially (e.g., because twice as many subjects were randomly assigned
to the treatment as to the control group).
Categorizing outcomes
• Tables are useful if there are several outcomes
of interest (e.g., death, myocardial infarction,
and hospitalization).
• Tables are also valuable when you wish to
categorize outcomes into different classes, such
as deaths due to cancer and deaths due to
cardiovascular disease, and especially if you
want to subcategorize them further, such as by
types of cancer.
Outcome
All cause
mortality
Vascular
Mortality
Non vas
Mortality
Major
coronary
event
Vitamins
14.1%
8.6%
5.5%
10.4%
Placebo
RRI
(95%CI)
NNH
13.5%
4%
(-3_12)
Not
significant
8.2%
5%
(-5_15)
Not
significant
5.3%
4%
(-8_17)
Not
significant
10.2%
2%
(-6_11)
Not
significant
Outcome
Stroke
Vitamins
5%
Revascular
10.3%
isation
Placebo
5%
10.6%
Cancer
7.8%
8%
RRR
(95%CI)
1%
(-12_13)
2%
(-6_10)
2%
(-8_11)
NNT
Not
significant
Not
significant
Not
significant
Thank you