Beruflich Dokumente
Kultur Dokumente
Sciences
WHY STUDY RESEARCH METHODS?
To learn about/predict behavior of people and objects
Must be able to sort/evaluate research/info others provide
Need to evaluate other programs/policies for success
Will make decisions (including lifestyle ones) based on research of others
Need to read/report research for classes
Research is about answering questions using the scientific
method
End Introduction
------------------------------------------------------------------------------------------------------------
CHAPTER 1
Research Paradigms
Quantitative research
Focuses on quantifying or counting variables; researcher is objective and removed from
involvement
Examples: The amount of marketing done by colleges; employers’ use of different
operating systems
Qualitative research
Focuses on beliefs/feelings; researcher deeply involved with participants
Examples are: Year long shadowing of executive to provide picture of that job;
unstructured interviews on experience in grad school to identify contributors to success
Qualitative requires more expertise in research methods in order to
prevent researcher’s bias contaminating results
Video
------------------------------------------------------------------------------------------------------------
Chapter 2
WAYS OF KNOWING
(these evolved into scientific method)
Traditionalism (authority based)
Rationalism (reasoning as source of knowledge)
Deductive reasoning (general to specific – all dogs have fleas,
Max is a dog, therefore Max has fleas)
Uses syllogism (general principle and specific assumption to
reach a specific conclusion)
Foundation of deductive reasoning is valid assumptions (Do all
dogs have fleas?)
WAYS OF KNOWING
(Falsification)
Empiricism (problems w/ valid assumptions so move to knowledge from
experience – observe events and draw conclusions)
Science is empirical (observation carries more weight than just reasoning)
Tries to escape making assumptions that may not hold true by checking beliefs against
observations
Unscientific empiricism – observation that is not systematically tested often leads to
erroneous conclusions or false generalizations
Falsification avoids the problems with causal reasoning
Maintains that no theory can ever be proved
Show beliefs are false by finding observations that disagree w/ belief
Test beliefs after they are created – never believe something is true – never say
something is true/proven (always subject to more testing)
System of Falsification
Statements to be tested must be falsifiable
Meaning there is some observation that would show them to be false
Cannot be inclusive (either it will rain or it won’t)
Cannot be definitions (all men are male)
Cannot be vapor statements (so vague that there is no specific prediction – it might rain
soon)
If a specific event shows the theory is not true, then it is falsified
If a specific even shows that it is, then it is confirmed or supported
Repeated attempts that result in no falsification lead us to believe the
theory is true, but it is never proven since the next observation might result in a
falsification
It is always possible that there was an error in methods that led to a wrong decision (so
only confirm or disconfirm, DO NOT PROVE)
Scientific Method
Science consists of repeated attempts to falsify theories
If falsified, then modified or replaced but if confirmed, then let the theory stand, but
still keep trying to falsify it
Scientific method involves these steps:
1. Observe some aspect of the universe.
2. Develop a tentative description, called a hypothesis, that is consistent with
what you have observed.
3. Use the hypothesis to make predictions.
4. Test those predictions by experiments or further observations and modify the
hypothesis in the light of your results.
5. Repeat steps 3 and 4 until there are no discrepancies between theory and
experiment and/or observation.
End Chapter 2
------------------------------------------------------------------------------------------------------------
CHAPTER 3
Ethics of Research
Beliefs about what is right/wrong
Questionable activities have greatest potential for ethical violations
APA Guidelines (1982) – no physical OR mental harm to
subjects (came from Zeller, Milgram studies)
Must get informed consent before subjects participate (they must be told
of risks and that participation is voluntary)
Anonymity v. confidentiality (subjects told in cover letter)
Buckley Amendment guarantees a right to privacy (not observing behavior
one would not expect to be observed)
At Tusculum, must complete Ethics in Research form (based on APA) and
have it signed by supervisor and instructor
COERCION OF SUBJECTS
(making people do something they don’t want to do)
Research carried out on the powerless which may be
indirect coercion
Students, homeless, prisoners, military
Once subjects start, they feel compelled to complete
(i.e. Orne study on adding digits)
Must be informed they can quit at any time
Researcher must stop them if they look uneasy
PLAGIARISM
(stealing another’s ideas and presenting them as yours)
A grave ethical lapse
Plagiarism has resulted in dismissal from school and
even jail time!
Biggest problems are
Using an idea in your report without giving full credit
Publishing the work of someone else as yours (Dr.
Hewish/Bell pulsar discovery – Nobel prize)
Fabricating data (book discusses Sir Cyril Burt, but recently his
records were released)
CHAPTER 4
Theory of Measurement
What is measurement?
Measure a property, usually with a number that tells something
about the property
Measurement is the relationship between a label and an object
Ruler tells the relationship between height (property) and a label (66
inches); if we know label, know height
All measurements tell the relationship between property of objects/events
and labels (intelligence is a property, while an IQ score is a label)
Measuring a
Person’s Intelligence
How intelligent they are
Whether the test scorer likes them
- How to overcome this problem? *
How they feel; not paying close attention or marking answer
sheet correctly; room is too hot
- How to overcome these problems? *
EVALUATING MEASURES
Good measures have little variability because they
reduce or eliminate both random errors and systematic
errors (biases or irrelevant values) and measure the true
value
The measurement is not the construct…just indicates it is there
Good measures start with operational definitions
Operational Definitions-Part 1
Specify for readers how a measurement was obtained
Give empirical observations collected to measure
construct
Researcher decides how to define the construct (For
example…measure intelligence by how quickly solve a
puzzle, or by GPA, or by annual salary)
Operational Definitions-Part 2
Constructs are subjective (different people measure them differently), but
operational definitions are objective (set rules so everyone measures the same
way)
If intelligence is defined as how large your annual salary is, there may be
disagreement about the definition, but if everyone measures annual salary the
same way, this eliminates variation in measuring
Good research relies on good operational variables, frequently by using
converging operations…several definitions converge on the construct so have
more faith are measuring the true value - For intelligence: measure puzzle
solution time, GPA, and annual salary
LEVELS OF MEASUREMENT
(Lower/Categorical)
Nominal (name only categories, no order)
Names as symbols of properties, no amounts…Numbers may be used, but only as ID
tags such as 1=male, 2=female
Math functions on these name tags make no sense (report frequency/percent, mode);
Examples: gender, race
Ordinal (ordered categories)
Specified order, but not distance
Tells quantity only relative to others (more/less)
Numbers only indicate relative position, not quantity so limited math such as
add/subtract
Simply ranks (report frequency/percent, median, range); Examples: Place in finishing in
race; arrival to class (first, second, etc.)
LEVELS OF MEASUREMENT
(Higher/Continuous)
Interval (real numbers, in order, according to quantity)
Zero point is arbitrary, can have negative numbers; Most math functions are okay, but
ratios don’t make sense
Report mean, median, mode, range, SD; Examples are temperature (C or F); test scores;
GPA
Ratio (same as interval, plus true zero)
Absolute zero indicates lack of the property, no negative numbers; All math functions are
appropriate, including ratios (twice as much, half as much)
Report mean, median, mode, range, SD; Examples are income (in dollars); number of
children
End Chapter 4
See Checkpoints
------------------------------------------------------------------------------------------------------------
Chapter 5
ISSUES IN MEASUREMENT
(Sampling)
What/who to measure depends on what you will do with the
results
Sampling
A population is every member of a well-defined group
A sample is a small group of examples chosen from the population for
measuring
Sampling techniques (or methods) are the procedures used to select the
sample out of the population
Generalization – the process of taking measurements
on samples and applying the results to the population
Generalizations
A sample is representative when measurements of it apply to
the population
The sample must come from the population to represent it
In 1936 Literary Digest wrongly predicted Landon over FDR (based on a
sample of its readers); Gallup (a newcomer) got it right (by using representative
random sample)
But in 1948 Gallup used old census data for sampling and predicted
Dewey as winner over Truman (this led him to refine his sampling techniques)
The sampling method is how you determine the
representativeness of the sample
SAMPLING METHODS
(Probability versus Non-Probability)
Non-probability means every population member does not have
an equal chance to be in the sample
These are NOT representative samples, so cannot generalize (cannot
determine probability of being selected for the sample)
Examples: Haphazard/convenience; Self-selection/ volunteers; Quota
sampling (Gallup in 1948); Snowball
Probability (random) sampling gives every population member
an equal and known chance or probability of being selected
SAMPLING METHODS
(Probability)
Simple random (numbers/names in a hat)
Requires list of everyone in the pop. as a sample frame
Can use a table of random numbers
Systematic or Nth (N = pop. size ÷ sample size)
Take a randomly ordered list, start at randomly chosen point
Stratified (divide population into strata/levels, select from each
strata)
Proportional or disproportional
Cluster (use existing groups/clusters as sample units)
For geographically dispersed pop. or when an experimental treatment can
only be applied to a group (multi-cluster too)
SAMPLE SIZE
Must be large enough to be representative
The more variation in the population, the larger the sample
needs to be
The more sensitive an effect that must be identified, the larger
it needs to be
Can calculate sample accuracy (e.g., ±3%)
Research/statistics books often provide tables
General guide is 10% of the population, with a
minimum of 30, a maximum of 1,000; but small populations
(<100) include everyone (census)
Types Of Measurement
Behavioral measures (observe behaviors)
Often only way to avoid social desirability bias
Some behaviors cannot be observed without changing the behavior
(guinea pig/Hawthorne effect)
Surveys or self observation (election/opinion polls,
marketing/attitude surveys)
Questionnaires designed around a single research question
Respondents must have and be willing to share the information the
researcher is after
Developing Surveys
Write many individual questions and pretest them (have others
read them and give feedback)
Must decide on open ended (easy to write, hard to analyze) versus closed
ended (pigeon hole responses) or combination of both
Once questionnaire is finished, have it reviewed for
facial/content validity
Pilot test questionnaire and procedures to collect data
See question examples, p. 107 and handouts
Survey Administration
In person, on telephone, by mail (less personal bias here)
Essential that all respondents be asked the same questions in same
manner to reduce variation
Do not reveal purpose of study in cover letter unless must (to avoid
acquiescence bias)
Non-response is a problem and must be addressed (essentially
self selection bias and those who feel most strongly are most likely to
respond)
Random non-response is okay, but systematic is a problem (best to get
two-thirds back, but at least half)
Can do follow-up calls (compare respondents to non-respondents)
Can compare respondents’ characteristics to known population factors
Interviews
Structured or unstructured
Often do a few unstructured and then develop
structured interview from results
Interviewing (in person/telephone)
Stick to standard 9am-9pm hours
Talk, look, act like those you are interviewing
Be polite and professional
Can tape interviews (ask permission and tell them they can
shut off recorder if desired – they often forget it’s there)
UNOBTRUSIVE RESEARCH
(those observed are not aware of it)
Archival research (existing databases/records)
Must establish accuracy of info
Must have info in the desired form (unit of measure)
Tusculum requires that you collect your own data
Content analysis (systematically analyze the content of recorded info)
Count the number of words/ideas (Pres. Clinton’s speech, entitled “The Era of Big
Government is Over” included 39 new government programs)
Content analysis of school board meetings showed the butterfly effect of random events
(no garbage collection)
Trace measures (deposits and withdrawals) – museum tiles
Hidden observations (beware of privacy laws) – urinal example
Systematic Observation
Decide beforehand what to observe and how to measure
Use check-off sheets
Define variables (ex: when measuring classroom misbehavior
record out-of-seat, talking, hitting, etc.)
Use one of these types of sampling
Continuous sampling (record all behaviors in a time interval such as an
hour…requires several observers)
Time point sampling (look & record, rest, look & record)
Time interval sampling (record all behaviors within a short time, such as 5
minutes)
End Chapter 5
See Checkpoints
------------------------------------------------------------------------------------------------------------
Chapter 6
Measuring and Using Correlation
(Essentially descriptive research, but goes further)
Correlation research focuses on relationships between
two variables
Used to describe and predict (but NOT explain…does not
establish cause and effect)
Finding relationships (co-relate means one variable
moves with another variable)
Measure two variables (x and y) on one set of subjects
Can be positive, negative, or no linear relationship
Correlations
Positive (direct) relationship means a high score on one
variable is associated with a high score on the other (and
vice versa); HI/HI & LO/LO
Negative (inverse) relationship means a high score on
one variable is associated with a low score on the other (and
vice versa); HI/LO & LO/HI
No linear relationship means no apparent linear pattern
Correlation Coefficients
Report DIRECTION and STRENGTH
For linear data, use Pearson Product Moment Correlation Coefficient
(ppm), r
Check scatterplot for linearity (other coefficients for curves)
Direction is indicated by the sign…a positive sign means a direct/positive
(HI/HI;LO/LO) relationship while a negative sign means inverse (HI/LO;LO/HI)
Strength is indicated by the absolute value of r , zero means no
relationship
r is from -1.0 → 0 →+1.0 (written r=.76 or r=-.68)
Example of positive relationship is grades and time studying, while
negative would be grades and missed classes
Making Predictions
If two things correlate, can use one to predict the other
The stronger the correlation (closer to ± 1.0), the more accurate the
prediction
Regression equations use a formula whereby a new y is predicted based
on a given x
If the prediction works, or is accurate, can use it even if there are
arguments of biased measures (SAT and grades)
Beware that relationships change over time and samples AND restricted
range in one variable causes problems (such as graduate school GPA)