2. What is ideal research?
Should be reproducible
Should withstand statistical analysis
Should test a theory / hypothesis / belief
Should be beneficial to the public
Should be systematic / empirical / critical /
Should have academic integrity
Should be publishable
Desist finding questions to your answers
4. Factors affecting confidence
interval
Sample size – Larger the sample size better is the confidence interval
Percentage – Represents the accuracy of the study
Population size – This is least important provided the samples are
randomly selected. This is important when the group is relatively
small and contains known group of people
5. Confidence level
This tells the researcher how confident the actual mean falls within the
Confident interval. Standard deviation if applied tells the researcher
How much variation that can be expected with the studied sample
Size. Ideal SD value is 0.5.
6. Sample size calculation
Confidence level = Z This is a constant value
90% - confidence interval - Z score = 1.645
95% - confidence interval - Z score = 1.96
99% - confidence interval - Z score = 2.326
Sample size = (Z-score)2 * Std Dev* (1- SD) / (margin of error)2
((1.96)² x .5(.5)) / (.05)²
(3.8416 x .25) / .0025
.9604 / .0025
384.16
8. Objective type
Physical characteristics
Testing universally applicable rules / laws
Testing hypothesis
Experiments
Surveys
Avoid the lure of numbers. Observation of researcher is more vital
9. Subjective type
Involves social life of groups
This study is usually conducted by observation and the findings
documented and explanations attempted for the observations
Usually social scientists use this modality
Always assume that your work will be scrutinised by the public
10. Types of objective study design
Descriptive
Analytical
Interventional
Greatest danger is not failure but non submission of your work
11. Descriptive study design
These studies consider variance of disease in respect of time, place
and person. Classic example of this design would be an attempted
study on the incidence of age related degree of progressive sensori
neural hearing loss.
These studies provide clues that can be used to design elaborate
analytical studies.
Two types of descriptive studies are possible i.e. cross sectional and
longitudinal.
12. Cross sectional study (Descriptive Design)
This study is based on single examination of cross section of
population performed at one point of time
Results can be projected on the whole population provided the
study is random in nature
This is a fast and inexpensive way of ascertaining incidence of a
disease
13. Cross sectional study - Steps
Objective of the study should be clearly defined
Population under study should also be defined clearly
Disease / health problem to be studied should also be defined
clearly (diagnostic criteria should be laid down)
Randomization of the sample should be ensured
Double blind trial has more validity
Make a list of variables
Prepare a questionnaire
Decide on a sample size
14. Longitudinal study (Descriptive design)
Observations are repeated in the same population over a
prolonged period of time by means of follow up examinations
Natural history of disease and its future outcome can be studied
Helps in identification of risk factors in disease causation
Also helps in finding out the incidence rate
15. Advantages of descriptive studies
Provides morbidity and mortality data
Provides clue to disease etiology
Generates hypothesis which can be tested by analytical studies
Provides data for planning, organizing and evaluating preventive
and curative services
Contributes to research in terms of disease occurrence by time
place or person
16. Analytical study design
Classic example of this design would be the study to ascertain odds
of developing noise induced hearing loss.
Intensity / duration of noise exposure should be factorized.
Age and sex of the patient (variables).
Analytical study design could be prospective and retrospective
17. Prospective study design (cohort /
longitudinal)
Difficult to perform
Tests the hypothesis obtained by descriptive study
Should proceed from cause to effect
This study is carried out on healthy people on whom exposure has
occurred and disease has not
Vulnerable groups should be followed over a period of time to
identify the risk factor
Costly to perform
19. Retrospective study (case control
study)
Easy to design and perform
This study is performed based on medical records
Study includes cases with health problems and controls without
disease
They should be matched evenly age for age and sex for sex to be
valid
Randomisation is a must
Cost of study affordable
20. Interventional study
Interventional studies attempt to demonstrate the cause-effect
relationships by altering the natural history of the disease by
intervention aimed at reducing the exposure to the offending
agent. (Sound in this case)
Control group should be included for comparison
Randomization should be followed to remove bias
Single / double blind protocol can be followed
21. Beware of variables
They should be identified correctly
Incorrect identification of variables will invalidate the entire research
Factors that could invalidate the entire research should be listed
and factorised
The trick is in trying to unearth surprising variables
22. List some of the variables in our
hypothetical project
Intensity of noise in decibel
Number of hours of exposure / day
Exposure of workers to ototoxic drugs
Surprising variable – temporary / permanent threshold shift
23. Common pitfalls
Sample size
Variables
Improperly formulated questionnaire
Improperly matched control
24. Types of sample
Convenient sample (ideally suited for our research scenario taken
up here)
Judgement sample (according to the one who is familiar with the
characteristics of the population under study)
Random sample (gives the most accurate and validated result)
25. Sample size
Don’t hesitate to take the help of statistician at this stage
For any successful research the confidence level should at least be
above 90% with error value of a minimum 5-10%
Avoid online sample calculators
26. Variables – dependent /
independent variables
All experiments contain variables at least one if not more
These can be measured / studied
Dependent variable – is dependent on independent variable
27. Categorical variables
Nominal variables – Can have two / more categories
Ordinal variable – can have two / more categories that can be
ranked
Dichotomous variable – can have only two categories (either or) like
male / female
29. data analysis
Attempt must be made to summarize the observed variables
If many variables are taken into consideration then coding and
categorization should be performed
Study of frequency distribution should be resorted to analyse
complex data
Data should be displayed as bar diagram / pie chart / histogram /
frequency distribution curves / x-y plots
30. Line graphs
Useful in tracking changes over a
period of time
Smaller changes are better
displayed
Can also be used to compare
changes over time even for more
than one group by changing the
colour of the line
31. Bar graphs
Can be used to compare things
between different groups
Can also be used to track
changes over course of time
This graph suits best if the changes
are larger
32. Pie charts
Best used when comparing parts
of a whole
Cannot be used to show changes
over a period of time
33. Area graphs
Similar to line graphs
Can be used to track changes
over time
Groups must be categorized
before displaying
34. X-y plot
Used to determine relationships
between two different things
X-axis is used to plot one variable
and the y-axis is used to plot the
other
If both variables increase at the
same time it is positive relationship
If one variable increases while the
other decreases it is negative
relationship
35. Mean / median / mode
Mean – is nothing but an average. It is the sum of values divided by
the number of values
Median is the value that divides the distribution into half
Mode is the value that occurs most often
36. Variance / standard deviation
This is the most preferred method of variation
It uses all the observation
Variations would be small if the observations are bunched closely
Variations if averaged will always be zero because positive
deviations away from the mean would cancel out the negative
deviations away from the mean
Squaring the average of deviations is resorted to, and this average
of squared value would always stay positive
Standard deviation is a measure of how spread out the numbers
are. It is actually the square root of variance and is indicated by
Greek letter sigma
39. calculation
Variance can be calculated by squaring the differences and
averaging them (21704)
Standard deviation is square root of variance = 147. This number
helps in comparison.
Use p values / chi-square test to test hypothesis
40. Before choosing a topic
Conduct feasibility study
Is it possible to complete within the given time frame
Affordability
Institutional support
Can you obtain necessary literature?
Will the topic be relevant after the completion?
41. Check list
Exact date of submission
Any word limitations
Intermediate deadlines to meet
Rules regarding the publication format
Tutorial support available
42. Points to be borne in mind
No harm should come to participants in the research (physical /
mental / social)
Children / elderly / mentally retarded should not be exploited
No physical / environmental damage should be caused
Anonymity / privacy should be ensured
Nothing should be done that would bring disrepute to the institution
43. Interviewer conduct
Friendly and formal
Schedule to be followed
Prior appointment to be sought
Treat all interviewees the same
Prompt don’t direct
Do not volunteer answers
Never be patronising
Be patient
44. Some useful research topics in
otology
Incidence of conductive deafness in children and their causes
Incidence of noise induced hearing loss
Measles infection – does it cause otosclerosis ?
Acceptability of hearing aids
Age related normal hearing in Indians
45. Title
Start off with a draft title
Keep polishing it
Avoid question marks in title
Include the period and place of study in the title if possible
46. aim
Here the aim of the study should be stated
Inclusion and exclusion criteria may be stated here as a
subheading (ideally done in materials and methods)
47. Introduction - chapter
Should contain an outline of your research
Should contain details of what prompted you to undertake the
study
It should also state concisely what you plan to do and where you
plan your work
Start writing this chapter first, edit it after completing the project
48. Literature review
This is central to all research
It informs the reader how well you have prepared for the topic
Here you take the opportunity to acknowledge other’s work
It also informs the reader the road you plan to take
49. Materials and methods
Here the exact research methodology followed is described
There should be a description of the tests used
Inclusion and exclusion criteria should be discussed in detail
50. result
Data should be presented
Data analysis should be presented here
Statistical tool used for the analysis should be discussed here
51. conclusion
Take time writing this one
Give your conclusions point by point in clear terms
Results should not be repeated but summarized here
Practical recommendations can be included here
52. Bibliography
List down all the references and citations
All references and citations should easily be identifiable
53. appendix
The material given here is for optional reading
Copy of questionnaire
Interview schedule
Copy of ethical committee approval
Copy of institutional approval
Editor's Notes
What I am not
Not a statistician
Not an epidemiologist
Ideally confidence interval should be 90 - 95%
It represents how accurate your sample matches the population studied.
Larger the sample size better is the confidence interval. This relationship is not linear.
95% confidence level means that your results. In lay terms it tells you the margin of error.
If the confidence interval is 50% then the result is questionable.
Sample size calculation for 95% confidence interval.
Margin of error is another constant = .05
This equation is for unknown population size / very large population size.
If the population size is small or known then calculators are available
Analysis of objectively observable data
Descriptive study design helps in formulation of idea for analytical / interventional studies.
This study describes the data being studied. It is also known as statistical research. It just gathers accurate data but does not go behind the scene to
Explain the cause.
Cross sectional study: Is a single examination performed on a cross section of population
At one point of time. Results can be projected across the whole population provided
It is performed in a random manner. These are fast inexpensive way of determining the incidence
Of a disease.
A small modification of this study could make it more useful if a series of cross sectional studies
Are performed at several points in time. This is known as serial survey design
Objective of the study should be SMART:
Specific
Measurable
Attainable
Realistic
Time bound
This design checks whether outcome is related to exposure.
Can be observational researcher does not influence / interventional (clinical drug trials)
Tests the hypothesis obtained by descriptive study or review of literature
Cohort – group of persons possessing a characteristic in common
Cohort is chosen from general population
Divided into 2 groups exposed and unexposed
Both groups should be evenly matched
Disease outcome and diagnostic criteria should be defined before hand
Follow up is similar for both groups
Exposed: with disease = a without disease = b
Unexposed: with disease = c without disease = d
Sample size a+b+c+d
Exposed incidence rate:a/a+b x100
Unexposed incidence rate: c/c+d x 100
Relative risk = a/a+b
______
c/c+d
Starts from a health condition and goes backwards through medical records.
This study is done to find a cause in diseased patients where the disease is rare with a long latency
Sparrow story
Teacher wants to know how students secure marks in maths test:
Dependent variable – test mark (0-100)
Independent variable – revision time / IQ of the student.
In experimental research the aim is usually to manipulate independent variables and then
Examine the effect this change has on dependent variable.
In non experimental research the researcher should not manipulate independent variables.
Eg. Study of behavioural changes after using narcotics.
Nominal variables – discrete and qualitative
Eg. Properties: types include House, shop condo etc.
Ordinal variable: qualitative ranking like good / average / bad
Interval variable – difference between 10-20 degrees C is 10
Difference between 40 – 50 degrees is also 10
Ratio variables are interval variables. It is a ratio of measurements. A distance of 10 m is equal to twice the distance of
5 m. (zero is clearly defined)
median – if observations are arranged in increasing order then the median is the middle of the observation
All values to be considered.
Mean height calculated and plotted.
Difference from mean is plotted