Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
0% found this document useful (0 votes)
34 views

Research Methods Powerpoint

This document provides an overview of research methods and the structure of a research project. It discusses the following key points in 3 sentences: 1) The course aims to develop skills in research planning, implementation, adaptation, and responding to changes through participatory lectures, group work, assignments including literature reviews and research proposals. 2) A research project consists of 4 components: assessing existing knowledge through literature reviews, generating and refining ideas, designing and executing studies to test ideas, and analyzing, interpreting and presenting resulting data. 3) Developing a clear research question is important, as the question forms the basis for data collection and should be focused enough to be answered within the project's scope while allowing for multiple possible

Uploaded by

Tenywaw Illiamu
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
34 views

Research Methods Powerpoint

This document provides an overview of research methods and the structure of a research project. It discusses the following key points in 3 sentences: 1) The course aims to develop skills in research planning, implementation, adaptation, and responding to changes through participatory lectures, group work, assignments including literature reviews and research proposals. 2) A research project consists of 4 components: assessing existing knowledge through literature reviews, generating and refining ideas, designing and executing studies to test ideas, and analyzing, interpreting and presenting resulting data. 3) Developing a clear research question is important, as the question forms the basis for data collection and should be focused enough to be answered within the project's scope while allowing for multiple possible

Uploaded by

Tenywaw Illiamu
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 95

Research Methods

Isabirye Moses
Aim
• develop skills in research planning for

– implementation of research projects,

– adapt to changes in ongoing projects,

– and respond to changing demands in the


course of one’s career
Course realization
• Participatory • Activity in lectures
matters
• Group / team work
– Shared responsibility
– Law of the minimum
• Collaboration
Course realization
• Lectures • Power point
presentation
• Power point lecture
notes
• Various articles
Course realization
• Assignments • Part of course work
(CW)
• Problems
• Information gathering
• Review of articles
• Contribute 50 % to
CW
Course realization
• Assignments • Literature review:

Examining the Climate


Change in the context
of DPSIR in Uganda

Research proposal
development
Course Assessment
• Continuous Assessment 40%

• Take home assignment: 20%

• Tests: 20%

• University Examination (3 hours): 60%


Course content
• The Research • Sample Surveys,
Enterprise Questionnaires and
• The Structure of a Interviews
Research Project • Sample Size
• Alternative Types of Calculations
Study Design • The Rationale of
• Experimental Design Research
• Quasi-Experimental • Critical review of
and Observational Literature
Studies
Course content
• Presenting and Reporting
• Styles of Data Analysis
• Statistical Models
• Types of Data Structure
• The Research Process
References
1. The Design of Research Studies – A Statistical Perspective, Part I: Planning
and Reporting, John Maindonald, Centre for Bioinformation Science, John
Curtin School of Medical Research and School of Mathematical Sciences,
Australian National University. john.maindonald@anu.edu.au
•  
2. The Rationale of Research, This is Chapter 8 from The Design of Research
Studies – A Statistical Perspective1, John Maindonald, Statistical Consulting
Unit of the Graduate School Australian National University.
• john.maindonald@anu.edu.au
•  
3. MSc Research Skills: Literature review, Citations, and List of references, D G
Rossiter, Version 1.01, ITC, Enschede, May 25, 2008
•  
4. Research,
http://en.wikipedia.org/wiki/Research_methods#Research_methods
Introduction
• Research
– Investigate
– Study
– Explore
– Delve into
– Examine
– Make inquiries
– Follow a line of investigation
– Look into
Introduction
• Components of a research activity
– generation of new ideas that may be worth
investigating.
• requires openness to new ideas.
– there must be critical scrutiny of all ideas,
whether they are an accepted part of
knowledge or new
Introduction
• Different types of studies call for different
approaches.
– Need for flexibility
• Account of scientific method must allow room for
the exercise of imaginative insight
• It must also pay regard to checks on the
unconstrained use of the imagination.
– To avoid myth and fiction that presents itself as
science.
Introduction
• Ideas may come in many ways:
– from working out the implications of
existing theory,
– in reverie (dreams),
– from one’s reading,
– from brainstorming sessions,
– as a by-product of the process of critical
scrutiny and testing,
Introduction
Logic Research and Experiment
Two kinds of problems based on system of
reasoning used in their solution
• Deductive Reasoning:
– From general principle to specific set of
conditions
• Inductive Reasoning:
– From specific to general
Introduction
Inductive reasoning
• Requires group of observations
– Often made under controlled conditions
• Factors studied vary by application of
treatments
• Other factors kept constant, we then have
an experiment
Introduction
• A typical problem facing a researcher: will the use
of a new treatment or different practice affect the
outcome of a phenomenon or enterprise and if so,
to what extent
• The problem cannot be answered with 100%
certainty.
• Therefore must also consider the risk and cost of
making incorrect decision
• An experiment is required to answer such a
problem.
Introduction
• What works for one person or for one
research project may not work for another.
• The origins of creativity are a deep
mystery, part of the mystery of our
humanness.
• Creativity has its best chance when the
research enterprise has captured the
imagination.
• Researchers who find their task boring
and uninspiring are unlikely to be very
creative.
Introduction
• When ideas appear, there must be
mechanisms for deciding which are worth
pursuing.
– How does one know which ideas are worth
pursuing?
• There can be no sure criteria.
– Staying away from past unfruitful research
may cause us to miss important insight.
Introduction
• However, some criteria can be used
– Repeatability
• It is appropriate to ask whether results can and
have been repeated, by different workers in
different places.
Introduction
• Planning should achieve a clear sense of
where research is headed, and of how it
will achieve its aims
• How does one get the data and do the
analyses needed for a convincing end
result?
Introduction
• In the design of data collection, and in
interpreting results, subject area insights should
mesh with statistical and data analysis insights
in ways that will vary from study to study.
• The researcher’s challenge is to put together all
the evidence from –
– the literature,
– the analysis of the researcher’s own data,
– less formal evidence that may not be amenable to
statistical analysis,
• in a manner that presents a coherent story.
Components of a Research
Project
• Four Components of a Research Project
– 1. assessment of the state of existing
knowledge;
– 2. generation and honing of ideas;
– 3. the design and execution of research that
will explore or test specific ideas;
– 4. analysis, interpretation and presentation of
the resulting data.
Review of literature
• What is the Current State of the Evidence?
• Researchers will be wise to attend closely to the
efforts of earlier researchers.
– Do literature review first to
• avoid re-inventing the wheel
• Avoid pursuing what is already known to be a bypath.
• Scrutinize all claims made by earlier researchers-
– Their methodology may be inadequate, or
– they may have misinterpreted their data.
• Learn lessons both in the successes of earlier workers and in
their mistakes.
Reviewing of literature
• What if the experts disagree?
• If all sides agree that there is as yet no definitive
evidence either way, and are taking different punts on
what the future may hold, that is healthy.
• Where both sides consider that the evidence supports
their judgment, the problem is clearly more fundamental.
– Underlying the disagreements are differences of opinion on what
is valid scientific evidence.
• It is then insightful to contrast the different sorts of
evidence on which the different protagonists rely
Reviewing literature
Interpreting Results
A Framework for Interpreting Results
• Scientific insight provides a framework within
which to interpret the statistical results:
– The data, and analyses, must be interpreted "in
context".
– look for possible biases.
– The methodology must be explicit
– Methods must be compared under similar conditions
Structure of a Research Project
Various Demands of Different Areas of
Research
• The Research Question
– The likely extent to which the researcher will
need to refine the research question varies
from one area to another.
– the research question must be clear, sharp
and largely un answered.
A good Research Question
• Start with a research question, or at least the beginnings
of a research question.
• Asking the Right Question
– basis for the planning of data collection
– Initially phrased in very general terms.
– Hone the research question into a clear focus.
– some refining of the research question can be done during the
preliminary stages of the research
– Avoid questions that are unclear, or that do not give the research
a clear focus, or that are too difficult to answer within the
project’s time and resource limitations.
A good research question
• Is the question scientifically well-posed, i.e. is it
stated in a hypothetical form that leads to a
research design and analysis with scientific
credibility?
• Does the research question require data that is
accessible or attainable at a reasonable cost or
effort?
• Is the research question posed in a way that
can explain variability, different outcomes under
different conditions?
A good research question
• Are the units of analysis (observation)
clearly identified?
• Is the question posed in a way that more
than one result is possible, i.e. the working
hypotheses can be refute?
• Does the research extend our
understanding of the phenomena being
investigated; does it elaborate, extend, or
fill in gaps in our present knowledge?
Structure of a research project
• Differences that may affect data collection, analysis and
interpretation include:
– 1. the methods that will be used for collecting data – experiment,
published data, data archives, cross-sectional or longitudinal survey,
etc.;
– 2. the extent to which you will need to develop new methodology or new
measuring instruments;
– 3. the extent to which validity seems an issue. Are the data what they
seem to be; do they really measure, for example, well-being? This is
commonly a key issue in marketing or health social science. It is often
an issue in biology. It is much less often an issue in physical science;
– 4. the signal to noise ratio – commonly low in marketing or health social
science and high in physics, with biology somewhere in between;
– 5. the types of measurement instrument – whether questionnaires,
visual assessment e.g. of a pattern on an agar plate, physical
measurement, or a mixture.
Structure of a research project
Eight Steps of a research project
• 1. Search out the research context
– Know the state of existing knowledge, what existing
data may be available, etc.
• What is the state of existing knowledge?
– talk to experts, and by reviewing the literature.
– look critically at claims made in the literature.
• may include
– (i) looking critically at the experimental or sampling design that
generated the data;
– (ii) critical examination of the data analysis;
– (iii) critical examination of the interpretation.
Structure of a research project
• 2. Canvass for ideas and formulate
specific questions
– Generation of a hypothesis is not a statistical
activity.
– Brainstorming techniques are often quite
helpful.
– Many different people may have light to shed
on the question at issue.
– It is necessary to talk to users, Focus groups’,
on which there is an extensive literature
Structure of a research project
• 3. Determine what type of study is needed
– The study may be an experiment, or a quasi-experimental study,
or a sample survey, or an observational study.
– what kind of study is most likely to provide good answers to your
research question.
– Here are some of the issues.
• i. Properly designed experiments allow clear cut answers. If
undertaken with proper care, there is often little room for argument.
• ii. It is not always easy to design an experiment so that results are
unequivocal.
• iii. Many important questions do not lend themselves to
experimentation. It is not ethical to expose different groups of
human subjects to different levels of radiation, in order to develop a
dose-response curve for the effects of radiation.
Structure of a research project
– iv. In addition to the logistical problems of doing experiments,
there are cost issues.
– v. Observational or quasi-observational studies are typically
much less expensive than experiments, and easier to mount.
– vi. Governments and organisations, by the changes they make,
are all the time carrying out experiments, though usually they
describe them as “reforms”.
– vii. In what is really a case-control study, every motor-cyclist and
every tenth car driver are stopped on a freeway and asked
whether they have had a serious accident, requiring hospital
admission, in the previous 12 months.
– viii. In all studies that have an observational element, there is a
potential for confounding. In a case-control study, the two groups
may differ in more than the exposure.
Structure of a research project
• 4. Design the study
• statistical issues to consider:
– For experiments,
• what are the treatment units,
• how large should they be, and how many of them
are needed?
• How can one avoid confounding?
• Would some form of blocking improve precision?
• How will information be collected on each
experimental unit (e.g. measure all plants, or just a
sample),
• and how should it be collected?
– For sample surveys:
• what is your target population?
• What sample design will give the best precision for
a given cost?
• How many primary sampling units are required,
• how many secondary sampling units, and so on?
• Will you design your own questionnaire, or will you
adapt an existing questionnaire?
• How can you avoid questions that may puzzle
respondents, loaded questions and/or ambiguous
questions. How will you handle non-response?
• design should include planning of the details of data
recording.
• Will you enter data onto a sheet, or directly into a
computer? If onto a sheet, do you need a specially
designed form or forms?
• If into a computer, do you need a computer entry
form that can be displayed on the screen. How can
you be sure that the data are entered correctly?
• In experimental work, photographs and/or video
recordings may be useful as records of information
that you may want to check on later.
Structure of a research project
5. Design and carry out a pilot study
• As a check of the study design, of
measurement devices or instruments, and
of practical aspects of data collection.
Structure of a research project
6. Carry out the study and collect the data
• Logistical, rather than statistical, skills are required at this point.
• Keep eyes and ears open for evidence of problems, or for the
unexpected.
• A factor that had not been incorporated into the design may turn out
to be important. There may be implications for your later
interpretation of the data.
• An adjunct to the process of data collection must be careful
checking and re-checking of data, to avoid errors.
• It is often helpful to do initial exploratory data summaries as data are
collected.
• Any problems in the data can be investigated there and then.
Structure of a research project
7. Analyse the data
• The data analysis has two parts:
– Exploratory data analysis:
• Examine various forms of data summary
– Formal data analysis directly addresses the issues
that the study was designed to examine.
• Following the formal analysis, there is further
exploratory data analysis that one can and
should do; to check whether there is anything
that the analysis has missed.
Statistical Science
• The Insights and Methods of Statistical
Science

– The role and nature of statistical science in


research
• Statistical science is the science of collecting,
organizing, analyzing and presenting data.

• Data collection, analysis and interpretation are


integral components of scientific research.
Statistical Science
The need for statistical evaluation
• Harvests from two equal fields cannot be the
same
• Differences result from genetic and
environmental differences beyond control
• These differences represent variability among
experimental units called experimental error
• Makes it difficult to compare treatments from
two units
Statistical Science
• The effect of the new trt is confounded with
unaccounted variability
• Thus an experiment with one replication provides a
very poor measure of treatment effect and also
• Provides no measure of experimental error
• Collection of experimental data will allow an
unbiased estimate of treatment effects & the
evaluation of treatment differences by tests of
significance based on measuring experimental error
Statistical Science
• Treatment effects are estimated by
applying treatments to at least two
experimental units and averaging the
results for each treatment
• Tests of significance assess the probability
that treatment differences could have
occurred by chance alone.
Statistical Science
Principles of experimental designs
• Replication: trt repeated two or more times.
– Provides an estimate of experimental error
– Provides a precise measure of trt effects
• Randomization: all units have an equal chance of
receiving a trt
– Assures unbiased estimates of trt means and
experimental error
• Local control (blocking): allows for certain restrictions
on randomization to reduce experimental error
Statistical Science
Research, Scientific method, and the experiment
• Research procedure:
– 1. Formulation of an hypothesis- a tentative explanation
or solution
– 2. Planning an experiment to objectively test the
hypothesis
– 3. Careful observation and collection of data from the
experiment
– 4. Interpretation of the experimental results- leads to
confirmation, rejection, or change of hypothesis
Statistical Science
Characteristics of well-planned experiment
• Simplicity: selection of trt and experimental
arrangement should be simple and consistent
with the objectives of the experiment
• Degree of precision: right design and enough
replicates ensure measurement of differences
at given degree of precision
• Absence of systematic error: treatments are
compared in a uniform environment
Statistical Science
• Range of validity of conclusions. Achieved
through replication in time and space, and
a factorial set of treatments.
• Calculation of degree of uncertainty.
Experiment design should enable
obtaining of results by chance alone.
Statistical Science
Steps in experimentation: for any type of
research
1. Definition of the problem. State it clearly
and formulate questions that give solutions
2. Statement of objectives. Can be in form
of:
– Questions to be answered
– Hypothesis to be tested
– The effects to be estimated
Statistical Science
Selection of treatments
• Selection of experimental material. Consider
objectives and population about which inferences
are to be made
• Selection of experimental design
• Selection of the unit for observation and the number
of replications
• Control of the effects of adjacent units on each
other
• Consideration of data to be collected.
Statistical Science
• Outlining statistical analysis and
summarization of results
• Conducting the experiment
• Analyzing data and interpreting results
• Preparation of a complete, readable, and
correct research report
• Statistical analysis is an important part of
experimentation. Statistics helps in designing
and evaluation of resulting numerical data.
Statistical Science
– The Design of Data Collection
• Faults in data collection.
– A bad design may give data that are incapable
of answering the question that is asked.
– Or undue effort may go into getting information
on features of the data that are irrelevant to the
question asked.
– Too small experiments are incapable of
providing answers to the questions that are
asked
– Too large experiments may give redundant data
that makes little difference to the accuracy of
the overall result
Statistical Science
There is a great deal more to statistics than p-values
• Statistics is not only about hypothesis testing and p-
values.
• Focus at developing a model that accurately describes
the data, aids in understanding what the data says, and
makes prediction possible.
– Every study should address clear focused questions.
– choose a few hypotheses that are to be tested.
– The questions that statistical analyses are designed to answer
can often be stated simply.
• Effective statistical analysis requires appropriate skills
gained from professional training in the use of modern
tools for data analysis, and from experience in using
those tools with a wide range of data sets.
Statistical Science
• The Data Analyst’s Tools
– Graphs
• One picture is worth ten thousand words [Frederick R.
Barnard, Printer’s Ink, 10 March 1927.]
• Generally, graphs are the best way to Scrutinize data and the
form of graph you draw matters.
– Every statistical analysis should be accompanied by graphs.
– Combine the computer’s ability to crunch numbers and present
graphs with the ability of a trained human eye to detect pattern.
– Statistics and Mathematics
• Statistics is not mathematics
• Statistical methods rely heavily on mathematical theory.
Statistical Science
• Statistical Software
• The interplay between computing power and theoretical
development has made a huge impact on statistical
methodology, both for design of data collection and for
analysis.
• These developments have taken advantage of the
increased power of computers and of the programs that
drive them.
• We can do a much better job on many analyses than
was possible ten years ago.
• None of the packages can be a substitute for expert
knowledge or assistance.
Statistical Science
• Why not use Excel for data analysis?
• Excel is a convenient tool for data entry, and
possibly for simple data checking.
– But take care:
• Excel will not object if you have spaces or non-numeric
values
• There will be no warning if wrong values are used.
• Excel’s statistical features have severe limitations and traps.

• Opt for more appropriate tools, that allow better


scope for your skills.
Structure of a research project
8. Write the report(s) and/or the paper(s)
• Statistical presentation is important.
• Communicate results clearly and
accurately.
• If those who need to assess or use the
results cannot understand the exposition,
the effort may have been largely wasted.
Structure of a research project
Effective Planning
• Planning should find a balance between thoroughness and attention to
detail on the one hand,
• and leaving room for learning as you go along. Here are points to keep in
mind as you try to strike the right balance:
– 1. Plan for review and re-evaluation after finishing one phase of your study and
before you move on to the next phase.
– 2. The results of the literature review may have big implications for planning. So
do not set plans in concrete until you know what the literature says.
– 3. Use a pilot study to test the design, the techniques and logistics before
proceeding with any major experimental or data collection exercise.
– 4. Think carefully about how you will handle changes to the plan that may be
forced on you by unexpected circumstances.
– 5. Plan your general approach to data analysis, and ensure that you will have
access to the resources and skills that you need.
– 6. Leave yourself room to be surprised by the unexpected!
Structure of a research project
The Literature Review
• Statistical questions that may arise in a review are:
– 1. Were there confounding factors; i.e. is it possible that the result is
explained by something other than the factor assumed responsible for
differences between groups?
– 2. Is the statistical analysis adequate? Is it correct?
– 3. Have the results of the statistical analysis been correctly interpreted?
• Another issue is experimental precision. Did the experimenter use
precise equipment, and / or a precise experimental design?
– a basis of good assessment of papers,
– and because of the implications for your own design and data analysis.
• A further issue is bias. Results may be highly repeatable, but they
may have consistent and unknown biases
Structure of a research project
• Designing the Data Collection
– talk to a statistician!
• You should ensure that the experiment
has sufficient accuracy that it will in
principle be able to detect effects of the
magnitude that are of interest.
Structure of a research project
Planning the Analysis
• See a statistician and plan out the broad
details of your analysis.
• You should get a sense of what general
style of analysis may be appropriate.
• At the same time, leave room for
messages, found in the data themselves,
about what analysis may be appropriate.
Structure of a research project
Be open to new, unexpected and interesting
insights!
• Careful thinking about the research
question, and planning, are important in
order to give structure to what you do.
observations that were not part of the plan.
Structure of a research project
The Ethics of Planning, Execution and Analysis
• Research must conform to accepted ethical principles. Avoid:
– Fraud, involving the faking of results or
– the manipulation of data or results, is obviously a serious breach of
ethical principles.
• Researchers who work with animals or with human subjects must
ordinarily seek ethical approval.
• Standards for medical research may include:
– Research must conform to generally accepted scientific principles.
– There should be a careful assessment of the relative risks and benefits.
– Published results should “preserve the accuracy of the results”.
– The protocol should include a statement of ethical considerations.
– There must be special caution where there may be environmental
effects.
• The quality of the science is an ethical issue. Flawed studies, if they
carry any credence at all, may mislead.
Alternative types of study design
• Structured methods for collecting data
include
– experiments, censuses or sample surveys,
– prospective or retrospective longitudinal
studies,
– case-control studies,
– cross-sectional studies,
– various forms of structured observational
study.
Alternative types of study design
The level of control exercised by observer
• Events controlled by observer
– Randomised experiment (with/without controls) x (with/without
replication) etc.
– Unrandomised experiment (includes haphazard assignment)
• (with/without controls) x (with/without replication) etc.
• Study of uncontrolled events
– Distinct Perturbation Occurs
• Intervention analysis
– Distinct perturbation usually not evident
• Domain of study restricted
– Assessment involving that restricted domain (i.e. not a random or other sample
from the whole domain of interest; sampling frame is not the whole of the target
population)
• Sampling over entire domain of interest
– Analytical sampling
– Descriptive sampling
– Sampling for Pattern
Experimental design
• A key feature of experimentation is the control
that the experimenter exercises over the way
that levels of the different factors combine to
affect the response.

• Simple experiments look at 1 factor at a time


• More than 1 factor is handled and interactions
between factors can be detected
Experimental design
• Experimental design issues
– Minimizing errors, variability, precision,
• Randomized controlled trials
– Treatments are randomly assigned to plots
• Randomized block design

Experimental design
• Principles of experimental design
– Blocking-
• treatment compared under similar conditions
– Replication
• Replication reduces variability and ensures that there will be
a valid estimate of experimental and other error
– Randomizing
• balance out the effects of factors that are not amenable to
experimental control.
Experimental Design
• Multiple Measurements on Each
Experimental Unit
• The Role of Experiments
• The Language of Experimental Design
• Multiple Levels of Variation – Blocks
• Confounding

Quasi-Experimental and
Observational Studies
• Data from quasi-experimental studies are
commonly analyzed as though they had
been gathered under experimental control.
– Data generation must mimic conditions of a
genuine experiment for statistical inferences
to be reliable.
Quasi-Experimental and
Observational Studies
• Some alternative types of non-experimental study
• Accidents of Nature or Human Behaviour
• Cohort Studies (Longitudinal studies; retrospective
studies; follow-up studies)
– some elements of control and some features of an observational
study exist
• Case-control studies
• Case-control studies aim, by the choice of 'cases', to
exercise an 'after the event' control that as far as
possible substitutes for direct experimental control.
• Cross-sectional Studies
– Essentially a cross-sectional study is a type of survey.
Quasi-Experimental and
Observational Studies
• Evidence for Bias in Non-Experimental Studies
– if trials do not follow accepted standards for randomization and
concealment, then biases will result.
– Non-experimental studies offer even greater opportunities for
bias.
• Experimental versus non-experimental studies
– Non-experimental studies are useful in drawing attention to
possible associations.
– Their results are in general compelling only when two or more of
the following conditions are satisfied
• (1) the conditions closely mimic those of an experiment or
• (2) the effect is large and has no other plausible explanation
• (3) there are multiple confirmatory sources of evidence.
Quasi-Experimental and
Observational Studies
• Knowledge Discovery in Databases (KDD)
• Before using databases:
– Know the sources of the data, and of the
purpose for which they were collected
Sample Surveys, Questionnaires
and Interviews
The Planning of Questionnaire Based Sample Surveys
• based on the purpose that the survey is intended to
serve
– identify the research question or questions,
– identify any relevant information that is available from
existing sources,
– establish that a questionnaire based sample survey is
the appropriate way
• Respondents are likely to be more co-operative if they
can be persuaded that the questionnaire
– addresses important concerns.
Sample Surveys,
Questionnaires and Interviews
• a preamble at the beginning should set out the purpose
clearly explaining how results will be used, and address
confidentiality issues.
• the survey is an intrusion on those who respond
therefore the planner has a duty to respondents to carry
out the task in a way that makes their effort worthwhile.
• When ready, sort out:
– logistical issues,
– sampling design issues,
– questionnaire design issues,
– Data analysis issues.
Sample Surveys, Questionnaires
and Interviews
Logistical Issues
• Depends on how responses will be made
• Interviews require skills
• Some form of motivation to respond may be required
• Surveys of official agencies, or of organizations, will
require the co-operation of the relevant officials
• Detailed logistics cannot be worked out until sampling
design issues are resolved.
• What are the different tasks that are involved?
• Who will perform these various tasks?
Sample Surveys, Questionnaires
and Interviews
Sampling Design
• 1. What is the target population, i.e. the population about
which information is required?
• 2. What is the sampling frame (Domain), i.e. the
population from which individuals will be sampled?
• 3. What method will be used for selecting the sample?
use a probabilistic sampling scheme to avoid bias = the
simplest is simple random sampling.
• 4. What steps will be taken to ensure high levels of
response?
• What efforts will be made to follow-up non-respondents?
Sample Surveys, Questionnaires
and Interviews
Developing the Questionnaire
• Use of existing questionnaires helps to make improvement and
saves on time and money
• Starting from scratch:
• 1. There are questionnaires that seek specific factual information.
– to discover how people spend their money, how much on food, how
much on sport, how much on entertainment, and so on.
• 2. There are questionnaires that investigate opinions, attitudes or
feelings. What is the attitude of year seven students to science?
• For item 1, the general nature of the questions is clear. The problem
is to express them in clear and unambiguous language.
• item 2 offers the greatest challenge. A good strategy must be
identified
Sample Surveys, Questionnaires
and Interviews
Steps for developing the questionnaire
• 1. Make a draft of the questionnaire. Check that it has a clear
coherent structure. Be sure to include a short preamble that
explains the purpose of the questionnaire, what will happen to the
results, what has been done to ensure confidentiality, and so on.
• 2. Get someone who is experienced with questionnaires to look over
it with a critical eye. Make any necessary revisions.
• 3. Seek the co-operation of 10-15 potential respondents. Administer
the questionnaire verbally.
• 4. This is a follow-up to step 3. Once each set of results is complete,
ask the respondent to explain their answers in a sentence or two.
(Probing).
• 5. Make any necessary revisions.
• Do a pilot survey (30 – 100 respondents; big survey).
• Do a summary analysis, there should be a review both of the
questionnaire and of the conduct of the survey.
Sample Surveys, Questionnaires
and Interviews
• The Analysis of Data from Sample
Surveys
Sample Surveys, Questionnaires
and Interviews
Sample survey design
• Stratified random sampling compared with cluster sampling
• Clusters and strata group together members of the population.
• 1. In stratified random sampling the sample frame is divided up into
relatively homogeneous strata. A random sample is then taken from
within each stratum.
• 2. In cluster sampling, the sampling frame is divided up into clusters,
often clusters of people who live in the same general locality. The
sampler then takes a random sample of clusters, though perhaps
making the probability that a cluster will be chosen proportional to
cluster size. For a given total sample size, cluster sampling
generally gives reduced precision.
Sample Surveys, Questionnaires
and Interviews
Combining stratified random sampling and cluster
sampling:
• (i) element sampling, (ii) cluster sampling, and (iii) multi-
stage sampling.
• These may be used individually, or combined, to provide
a sampling system.
• The sample scheme determines how sample elements
or clusters are chosen.
• Options are:
– simple random sampling, and stratified sampling, etc
• In multi-stage sampling this choice is available at each
stage.
Sample Surveys, Questionnaires
and Interviews
Multi-stage sampling
• The sample procedure may be multi-layered, leading to
multi-stage sampling.
• At each stage the method used may be stratified random
sampling, or cluster sampling, or a mixture of the two.
Stratified Random Sampling – The Choice of Strata
• For stratification to be effective one needs a variable,
positively correlated with the outcome that is of interest,
that can be used to define the strata
Sample Surveys, Questionnaires
and Interviews
Questionnaire Design
• Research questions must translate into a
set of questions, and into a questionnaire,
that will provide answers to the questions
to which you as a researcher want
answers.
• What steps will help ensure responses
that will give reliable and valid answers to
the research question?
Sample Surveys, Questionnaires
and Interviews
Questionnaires as Instruments
• a particular form of questionnaire may be
refined to the point where it becomes a
recognized social science "instrument'',
widely used by different researchers.
• A key issue is: "What does the instrument
actually measure?'‘
Sample Surveys, Questionnaires
and Interviews
Qualitative Research
• Gives substantial insight into the likely social
dynamics required before mounting a
quantitative study.
• Qualitative studies may be treated as complete
in themselves, or they may be explicitly intended
to complement a quantitative study.
• Qualitative studies may be used to generate
questions that can then be addressed in follow-
up quantitative studies.
Sample Size
Relevance
• 1. Researchers should certainly have a rationale for the
size of their study. Size, i.e. number of replicates
• 2. Where improvements in study design allow improved
precision, this is usually preferable to increasing the
sample size.
• 3. Each new study should be seen as part of a total
learning process.
• 4. In highly exploratory studies, the effort put into trying
to get high precision may be largely wasted.
• 5. Sample size is highly important in stand-alone
compared to experiments in series.
Sample Size
• 8. The aim should be accurate estimation
of variability, and ensuring there are
enough degrees of freedom to do this.
Sample Size
Information Required for Sample Size Calculations
• Information that may be more widely relevant to
understanding the scientific context and to the
design of data collection. For comparing
proportions, a conservative (i.e. erring on the
large side) estimate is obtained by assuming
that the population proportion is 0.5.
Sample Structure
The Rationale of Research
• Data
– Basis for scientific explanation to complex
facts
• Theory
– Helps to select relevant data to collect
• Models
– Mathematical and non-mathematical
– The formula is a model, not the real thing.
– It is not totally accurate
The Rationale of Research
Regularities (Law-Like Behaviour)
• Mathematical models describe law-like
behaviour, i.e. one can use the model to
describe or predict.
• Statistical model components
– Signal
– Noise
• Minimising noise
– 1. By using more accurate measuring equipment.
– 2. By improving the design of the data collection.
Rationale of science
• Science as hypothesis testing
• Science as a Human Activity
– Fallible Scientists
– Dominant authorities
– The Logic of Science and the Sociology of Scientific
Communities
– Reductionist Scientists?
– Commercial Pressures
– The Uses of Controversy
A successful proposal

You might also like