The Research Process
Jan 02, 2016
The Research Process
Parts of The Research Process
Part 1 The research environment Part 2 The process of conducting
research Part 3 Professional and personal research
proposal writing Part 4 Research project management –
theory and practice
Part 1 The Research Environment
What is research?From Dictionary.com
1. Scholarly or scientific investigation or inquiry. See Synonyms at inquiry.
2. Close, careful study.
Personal research– For your own benefit
Scholarly research– For the benefit of society
Why do research?
What do you think???
Thesis (hypothesis)
Definition for Dictionary.com – 4 definitions1. A proposition that is maintained by argument.
2. A dissertation advancing an original point of view as a result of research, especially as a requirement for an academic degree.
3. A hypothetical proposition, especially one put forth without proof.
4. The first stage of the Hegelian dialectic process.
Research Objective
Provide evidence to prove or disprove a thesis (hypothesis)
Thesis – drivers will reduce their speed if we post their speed on a variable message sign
Provide evidence to allow the development of a thesis (hypothesis)
Thesis – we have observed the intersection crash rates at intersections with and without all-red clearance intervals and crash rates remain the same – all red clearance intervals do not impact crash rates.
Importance of Objectivity
Disproving your hypothesis can be just as important as disproving your hypothesis– Some motorists speed up
to see how fast they can go when their speed is posted to a VMS.
Research Ethics
Need for openness– Sharing results is essential
Mechanisms to promote openness– Protection of intellectual property
Patent protection Copyrights Speedy publication
– Job promotion and notoriety
Mechanisms For Openness
Promote publication of results Impose a moral duty to publish research
findings Creates an environment for sharing findings Establishes importance to taking ownership
(credit) and accountability for findings
Honesty in Reporting Results
Report results even though it may conflict with your hypothesis
Data should be thought to be valid unless determined otherwise
All research work should be reviewed by peers Wise sponsors will have the results of one researcher
reviewed by a peer researcher rather than using the results blindly
Deceptive practices can result in the loss of a job and loss of reputation
Errors, Negligence, and Deception
Errors – honest mistakes– Acknowledge at the earliest convenience
Negligence – Omission or commission of scientific process
– Did not check for normality of error terms
Deception– Fabrication– Falsification– Plagierism
What is an outlier?
Residual plot
VMT
California-3
-2
-1
0
1
2
3
4
0 100,000 200,000 300,000 400,000
Florida
Texas
New York
CaliforniaFat
al C
rash
Rat
e S
tand
ard
Dev
iati
on
VMT
Are Florida or Texas outliers?
Outlier
If kept these data have tremendous impact on you statistical results
If they are thrown out of the analysis they should represent anomalous data
Rule of Thumb - Should be a minimum of two or three standard deviations away from the mean
Regression, inspect results
2 4 6 8 10 12 14
X-Axis
0
5
10
15
Y-A
xis
R-square = 0.748 # pts = 13 y = 0.578 + 0.352x
Title
Coefficients Standard Error t Stat P-valueIntercept 0.431690739 0.467807938 0.922795 0.377842
4 0.374758631 0.075277665 4.978351 0.000555
Inspection
One data point was added
0 5 10 15
X-Axis
0
5
10
15
Y-A
xis
R-square = 0.479 # pts = 14 y = 10.4 + -1.24x
Title
Coefficients Standard Error t Stat P-valueIntercept 11.67086557 2.145346542 5.440084 0.000204
4 -1.404568306 0.358929371 -3.91322 0.002421
Peer review process
Assumption – The peer community is the most capable of reviewing a peers work
For a peer review process to work it has be blind – Double blind is preferable
Where is Peer Review Process Applied
Most journal articles Proposals to academic research program
– NSF
US DOT programs are reviewed by professional peers Research findings and reports are always reviewed by
professional peers and sometimes academic peers Dossier for academic promotion and tenure
Publication Reviews Types
Highly academic journals conduct double blind reviews Journals like the Transportation Research Record do a
single blind review by professional peers Academic and scholarly conferences with conference
proceedings review the full paper like the TRB annual meeting
Professional conferences like ITS America’s annual meeting and ITE’s annual meeting only review abstracts
Credit for Work and Accountability
Forms of credit– Authorship– Acknowledgement– Citation
ACKNOWLEDGMENTSThis research was sponsored by a Transportation Statistics Research Grant from the Bureau of
Transportation Statistics, United States Department of Transportation and was performed at the Connecticut Transportation Institute in the University of Connecticut. The authors thank Yusuf Mohamedshah of Federal Highway Administration for providing us with the HSIS dataset and assisting us in interpreting it. The authors also gratefully acknowledge contribution to this project made by Dr. Don Tepas and Jianhe Du. Thanks also go to Dr. Diane Lambert of Bell Labs for providing ZIP model computer codes.
Taken From: Xiao Qin , John N. Ivan, and Junfeng Liu, “A Hierarchical Bayesian Estimation of Non-linear Safety Performance Functions for Two-Lane Highways Using MCMC Modeling” TRB annual conference, 2003
Credit for Work and Accountability, continued
Authorship– The senior author should have provided the most
intellectual contribution and direction to the paper or report
To be named an author, the individual should– Provide an intellectual contribution or – Contribute direct and substantively to the work
Credit For Work
Citations– Always cite some one else’s work that has provided
supporting data for your research– If the prior researchers contribution was significant,
reference them by name in your article or report– Provide page numbers for quotes or citing of facts
Conflicts of Interest
A conflict between a person's private interests and public obligations
Actual conflict of interest A perceived conflict of interest Positive conflict of interest
A conflict of commitment Disclosure of conflict to sponsor The engineering and business worlds are rife
with conflicts of interest
Part 2 – The Process of Conducting Research
Usually we are presented with a topic– Improve safety on rural expressways using low cost and conventional
approaches Some time we are given general areas of interest and asked
to come up with innovative research ideas– Research topics must target innovative strategies for
exploiting state-of-the-art information and information/communication technologies in surface transportation systems to meet the challenges and constraints associated with mobility.
In frequently a researcher will get the opportunity to do exactly what they would personally like to do – study something for its intrinsic value.
Idealized research process steps
Identify a problem Conduct a review of the practice and the
literature Define the problem – problems statement Define a methodology to address the problem Execute you research plan Report findings Make recommendation
Getting started
Identify a topic area of interest (to you, society, or sponsors)
Define the problem. – Sometimes the objective and purpose are clear– Most of the time considerable background
information must understood to frame the problem Albert Einstein went so far as to say that "the formulation of a
problem is often more essential than its solution"!
Once you have an idea
Go through a vetting process– Talk to your colleagues about your idea– Consult experts– Consult practitioners
Vetting can help to toss out ill-conceived topics and refine good ones
Knowledge is cumulative
Next the researcher must review what has been done before– Perform a literature review– Identify gaps in the literature
Further define the problem– Provides a theoretical rationale of the problem
Define the problem
What are the specific issues that need to be addressed Define how your work contributes (rather than
duplicates) what is known Define how your work will relate to existing theory
regarding the issue Qualifying the risk involved in the research
– You may not find what you expect
Identifying the possible and likely out-comes
Types of problems
Descriptive.When a study is designed primarily to describe what is going on or what exists.
Relational.When a study is designed to look at the relationships between two or more variables.
Causal.When a study is designed to determine whether one or more variables causes or affects one or more outcome variables.
Scholarly Research
The researcher must be knowledgeable of the international state-of-the-art and know of the leading experts.
Must advance the state-of-the-art – create new knowledge
Must employ state-of-the-art research methods Generally involves some primary data
collection
Structuring of research problem statement
Methodological approach– The nature of the problem forces the approach
taken
Problem solving approach– Knowledge of the problem (known theory) forces
methodological approach
Research Method (methodological view point)
Washington identifies three research methods– Observational methods
Involves recorded or recording observations according to a plan– Example, the development of car following theory
– Experimental methods Changing one or more variables and observing the response
– Example, determining speed reduction due to speed calming technologies in the work zone
– Surveys Collecting information from subjects that is normally not available
– Example, collecting travel diaries to better understand trip chaining behavior
Steps in Research (taken from Washington)
Theory
Theory
Maze’s addition
Experimental Approach
Most commonly used Involves stating a hypothesis as part of the
problem definition Suppose that our original research issue is that
speed display monitors will make work zones more safe – how do we establish a research hypothesis to measure this?
Null Hypothesis
We either accept or reject the null The null and nil hypothesis are sometimes the
same– For example – when testing whether regression
parameters are statistically different than zero – the null and the nil are the same.
The null hypothesis should probably be what you feel you are most likely to nullify in the test.
Hypothesis statements
Hypothesis – we expect that the speed display monitor will result in statistically significant reduction in speed.– The null hypothesis is that speeds remains the
same (assume this means the mean speed)
The alternative hypothesis – The alternative hypothesis is that speed was
reduced
Typical Hypothesis Test Illustrates the Issues
Typically a test is performed to determine if there is a statistically significant difference in
– The mean value– The variance– The distribution
Level of significances is the area in the tails of the distribution
Formulated decision rules and set levels for the probability of error
Speed
Speed DistributionAfter ImplementingA Speed Display Monitor
Speed DistributionBefore ImplementingA Speed DisplayMonitor
Type I and II errors
lies in lies in acceptance interval rejection interval
Accept the No error Type II errorClaim
Reject the Type I error No errorclaim
Decision rules and set levels for the probability of error
AcceptReject
Area where we incorrectly reject Type I error, referred to as significance level)
Area where we incorrectly acceptType II error, referred to as
Issues
The way the hypothesis is stated is related to the potential for error
Much emphasis is placed on the statistical significant of a test and not on
When a test is stated, the research needs to understand the trade-off between level of statistical significances and
Problem solving approach (William M. Trochim )
Deductive reasoning works from the more general to the more specific. Sometimes this is informally called a top-down approach.
Inductive reasoning works the other way, moving from specific observations to broader generalizations and theories
Problem Solving approaches
Deductive Approach Inductive Approach
Deductive or experimental approach
Quantitative research– Involves theorizing or conceptualizing a relationship
(usually some kind of model) First step is to understand causal relationships Second step is to specify model (requires understanding of
the nature of the relationship)
– Empirical analysis affirms or disaffirms theory
Inductive approach
Quantitative research– Collecting the universe of variables– Conducting exhaustive search for relationships
First step – identifying statistical relationships Second step – theorizing causal relationships
– Empirical analysis is used to create theory
Deductive approach
Developing a theory– Develop a causal structure and order– Example – What causes people to travel
Demand for consumption (food, goods, life style) Work force participation – Travel to work Gather material – Travel for shopping Travel to life style enhancing experiences – Travel for
recreational and social events
Causal ordering
Demand to Consume
Work
Shopping
Social and Recreational
Trip Making
Demand toConsume
Work
Shopping
Social and Recreational
Trip Making
Why is causality importantThe rate of stork sightings and births increased in northern Europe following the end of WWII. What is the causation?
Couple has Baby
Stork’s feel flying is safeOver northern Europe
Storks return to Northern Europe
End of WWII
Male Soldiers return from War
Soldiers return to wife or girlfriend
Bombing in NorthernEurope ceases
Cau
sati
on
No Causation
Aggregation Fallacies
Caused by not knowing causality– Ecological Fallacy– Individual Fallacy– Selective Fallacy– Universal Fallacy– Contextual Fallacy– Cross-sectional Fallacy– Historical Fallacy
Fallacy Definitions – importance of understanding causation
Ecological Fallacy – Falsely inferring that what is true of the group must
be true of the individual Rich people make more trips therefore rich individuals
must like to spend more time in travel
Individualistic Fallacy– Falsely inferring that what is true of the individual is
true of the group Since I am Swedish and I am tall, all Swedes are tall
Fallacy Definitions Continued
Contextual Fallacy– What is valid for one group at one time is valid at another time
or under a different context It was a good idea to regulate commercial transportation in the
late 1890s and, therefore, reregulation is good idea today
Universal Fallacy– What is valid for the universe of individuals is valid for specific
groups The metropolitan area of Des Moines will benefit from the
reconstruction of I-235, therefore it is good for downtown Des Moines.
Data Fallacy continue
Cross-sectional Fallacy– Falsely extending one-time observation to all times
During the 1990s, tech stocks were the right vehicle for investment and therefore you should stick your money in tech stocks