Statistical Analysis An
Overview Introduction
Dr. AJIT SAHAI
Faculty- Consultant (Research)
Ex Director – Professor Biometrics
JIPMER, Pondicherry
Experimental studies
Clinical trials
Laboratory trials
Community trials
(a) Clinical trials
1. Controlled trials
parallel or concurrent controls
a. randomized
b. non-randomized
sequential controls
a. self-control
b. cross-over
external controls ( including
historical )
[Link] trials
B. Laboratory/experimental designs
Completely randomized
Randomized block
Latin square
Lattice
Cubic
Rectangular
Split plot
Factorial
Common Pitfalls in Research Protocols
Statistical Methods
Dr. AJIT SAHAI
PhD, FIAPSM, FIPHA, FSMS, FRSS
Faculty-Consultant (Research/Biostatistics)
KGMU, Lucknow.
Formerly-Professor & Head,
Dept of Medical Biometrics, JIPMER, Pondicherry.
A few generalized aspects
Title
Research Question(s)
Hypothesis(es)
Aims Objectives(Primary/Secondary)
Tools; Techniques & Methods:
Case definition (Controls)
Eligibility for recruitments
Methodology Aspects
Study design
Standardizations of tools; techniques of
measurements
Scoring & grading systems
Sampling & Sample
Random selections & allocations
Statistical tests proposed
Results-Discussions-Conclusions
and Recommendation Aspects
Tables; Graphs/Charts and Write-ups
Reveal the facts –Scientific Strengths
Never Censor the weaknesses (- ves)
Never discuss and conclude without
evidence or beyond study results
Never construct recommendations
beyond the study domain or scope
WHY? Sampling
A study of entire population is
impossible in most of the situations
Sometimes, the study process
destroys (animal sacrifice) or depletes
the item being studied
In such situations the only alternative
is sample study
Advantages
Sample results are often more
accurate;
Apart from being
quick and less expensive
Generalization strength -
Validity of Results
Target population
A common sense judgment is required to
make inferences from the sampled to the
target population
Sampled population
Random sampling is required to make
statistical inferences from the sample
to the sampled population.
Sample
Sampling Error (E)
Sampling Error
(E)
Sample size (n)
Sample-Size Vs. Sampling-Error
The sample size may have to be
directly proportional to the
heterogeneity in the population,
Whereas, the sampling error is
always inversely proportional to
the sample size.
Non-sampling error
Non-response
Response
Non-coverage
Measurement
Probability Sampling
The techniques of sampling may be
classified as;
“Probability sampling” such as;
- Simple random sampling,
- Stratified, cluster, systematic,
- Multi-stage and multi-phase
sampling; whereas,
Sample Size - Methods Vs Studies
Exploratory studies Confirmatory studies
Experimental studies Observational studies
Clinical studies Population studies
Cross-sectional studies Longitudinal studies
Prospective studies Retrospective studies
Cohort studies Case-referent studies
GENRALIZATION
If samples are properly selected,
probability methods can be used
to estimate the error in the
resulting statistics
It is this aspect of sampling that
permits investigators to make
probability statements about the
observations in a study
Power of a Study
It is not only the sampling method but
the sample-size also equally responsible
for the power of a study
For this reason, investigators must plan
the sample size appropriate for their
study prior to beginning research
Sampling methods
Random sampling
Simple Random Sampling (SRS)
Stratified Random Sampling
Systematic random sampling
Ratio/Regression methods
Systematic sampling
Cluster sampling
Multi-stage sampling
Multi-phase sampling
Deficiencies in published papers
no detail on sample size 89%
no confidence intervals 86%
inadequate detail on randomisation 60%
no details of concealment 77%
Moher, D. CONSORT: an evolving tool to help improve the quality of reports of RCTs.
JAMA, 1998, 279: 1489-91
defects in 18-68% of abstracts
Pitkin, RM et al, JAMA 1999, 281: 1110-1
errors in 4-67% (median 36%) of citations
Wager, E & Middleton, P: PIRATES abstract submitted to
Peer Review Congress, Barcelona, Sept 2001
Common Errors When Reporting
RESULTS
Errors of Omission:
Not accounting for all study subjects
Not naming which statistical tests were
used for specific analyses
Not presenting the results in a clinically
relevant units
Errors in the Analysis
Lack of power
Failure to adjust for multiple comparisons
Analysis by treatment received and not by
intention to treat
Errors in Interpretation
Not recognizing the limits/meaning of p
Data Analyses/Syntheses Process
Screening /Verification/Cleaning up
Organization
Compilation /Collation &
Classification (Nominal/Ordinal/Numerical)
Grouped Frequency Distribution
Tabulation
Graphical representations
Summary Statistics
Rates, Ratios and Proportions
Incidence & Prevalence (Point /Period)
Numerator and Denominator
Per – Cent / Thousand / Lakh / Million
Summary Statistics
Measures of
Central Tendency – averages
Dispersion/Deviation
/Scatter/Variation/ Spread
Location –percentiles /quintiles
The role of Central Tendency
and Deviation
-3s -2s +2s +3s
Mean
95.5%
99.7%
Biological Observations
Though this universe is full of
uncertainty and variability,
a large set of experimental /
biological observations always
tend towards a
Normal distribution.
Inferential Statistics
This unique behavior of
data
is the key to entire
inferential statistics.
Population Probability
Distributions
such as;
Normal
Binomial
Poisson
Rectangular
Sampling Distributions
Like;
Chi-square ‘X2’
Student’s
‘t’
and ‘F’
Population & Sampling
Distributions
frequently used for
probability calculations and
also for
testing the hypotheses
through various tests of
significance
Possible Conditions of Null
Hypothesis
H0 is TRUE H1 is TRUE
REJECT H0 TYPE I Correct
error decision
ACCEPT H0 Correct TYPE II
decision error
Probability (P) - Error types
The size of type I error is the probability
that the sample point will fall in the critical
region when H0 is true (α)
The size of the type II error is the
probability that the sample point will fall in
the non-critical region when H1is true (β)
Guidelines for Writing Results -
The Study as it was Conducted
Specify the dates of the study
Provide a schematic summary
Describe the characteristics of each group
Indicate if the sample is representative
Indicate if randomization was successful
Describe duration and nature of follow up
For observations based on judgment,
provide assessment of consistency
Guidelines for Writing Results:
The Study Outcomes
Present the results for all primary endpoints
Report 95% CI
Report actual p values for all primary analyses
Report the main findings in figures or tables
Report statistical findings in detail
Report confounders
Guidelines for Writing Results:
The Study Outcomes
Indicate compliance and deviations from
the protocol
Report treatment related SE/AE
Describe the treatment of outliers
Account for all observations; explain
missing data
Report anecdotal evidence or
observations that might influence
interpretation
Good publication practice
Read “Guidelines on good publication practice”
from the Committee on Publication Ethics
(COPE)
Available free at [Link] [Link]
COPE guidelines on good
publication practice cover
Study design and ethical approval
Data analysis
Authorship
Conflicts of interest
Peer review
Redundant publication
Plagiarism
Duties of editors
Media relations
Advertising
Dealing with misconduct, including sanctions
Why publish?
Because you have something important to say
To change practice
To promote thought or debate
To allow examination of your work
“Fame and the love of beautiful women”
Money
Career advancement
To entertain/divert/amuse
To educate
To console
What do editors want?
Excitement/ “wow”
Importance
Originality
Relevance to the audience
True
Clearly written
Engagingly written
Utility of information
Utility = relevance x validity x interactivity
work to access
The basics of writing a paper
1. Before you begin
What do I have to say?
Is it worth saying?
What is the right format for the message?
What is the audience for the message?
Where should I publish the message?
How can I best use paper and the web?
The basics of writing a paper
2. The importance of structure
Structure is everything.
The basics of writing a paper
3. Possible structures
Beginning, middle, end (AA beginning, a muddle, and an end.@
Philip Larkin)
Tell people what you are going to say, say it, tell them what
you=ve said
Rudyard Kipling: (AI keep six honest serving me, (They taught
me all I knew), Their names are What and Why and When, And
How and Where and Who?)
The basics of writing a paper
3. Possible structures continued
News story: (Story in the title; story in the first line; expand
slightly on the story in the first paragraph; give the evidence for
the story; give the counter view)
ImraD (Introduction, methods, results, and discussion)
Stream of consciousness
Chronology: diary, autobiography
The basics of writing a paper
3. Possible structures continued
A list
Something very formal: for example, sonnet, limerick, haiku
Ulysses
The basics of writing a paper
4. Still structure
Make sure that readers know where
they are, where they are going, and
why.
The basics of writing a paper
5. IMRaD
Introduction--Why did I do it?
Methods--What did I do?
Results--What did I find?
Discussion-- What might it mean? What is our overall finding? What
are the strengths and weaknesses of the study in relation to other studies?
Why might we have got different results? What might the study mean,
particularly for clinicians or policy makers? What questions remain
unanswered and what next?
The basics of writing a paper
5. IMRaD (Introduction)
Why did we start?
What has gone before - ? A systematic review
Why was this study needed?
Be sure that readers understand the importance of the
study-but don=t overdo it short
Don=t try to show readers that you have read
everything
Short, short, short
The basics of writing a paper
5. IMRaD (Methods)
Like a recipe
For informed readers this is the most important section
Describe how subjects were selected and excluded
Don=t describe standard methods in detail - use references
Statistics
Ethics
Remember that you can put more detailed methods on the web--
for example, questionnaire
The basics of writing a paper
5. IMRaD (Results)
Stick to what is relevant
Be sure to include basic descriptive data
The text should tell the story
The tables give the evidence
The figures illustrate the highlights
Don=t include just percentages or p values
Include confidence intervals
Think about absolute risk, number needed to treat, etc
Avoid beginning to discuss the implications or strengths and weaknesses of your
study
The basics of writing a paper
5. IMRaD (Discussion)
Statement of principal findings
Strengths and weaknesses of the study
Strengths and weaknesses in relation to other studies, discussing
particularly any differences in results
Meaning of the study: possible mechanisms and implications for
clinicians or policymakers
Unanswered questions and future research
Go easy on the last two
The basics of writing a paper
Topping and tailing
Title: Include design; Don=t try to be clever
Abstract: Must be structured; Include some
numbers, not all
References: Keep to the essentials
Covering letter: Something very crisp
Authorship, acknowledgements, competing
interests