download

Matakuliah
Tahun
: D0064 - Sosiologi dan Psikologi Industri
: Sep-2009
RESEARCH METHODS IN INDUSTRIAL
PSYCHOLOGY & ORGANIZATION
Pertemuan 05 - 06
RESEARCH QUESTIONS
• Every study begins with a research question
• Research question can be general or specific
“What causes people to like or dislike their
jobs?”  general
• To be useful the question should specify exactly
what is being studied. A better question that is
more specific is:
“Does level of pay effect how much people like
their jobs?  specific
Bina Nusantara University
3
IMPORTANT RESEARCH DESIGN
CONCEPTS
VARIABLES:
The basic building blocks of a design. A variable is
an ATTRIBUTE of characteristic of people or
things that can vary (take on different value)
Common variables in organization research:
People’s ability: intelligence
Attitude: job satisfaction
Behavior: absence from work
Job performance: weekly sales
Bina Nusantara University
4
Variebles are quantified so that they can be
analyzed with STATISTICAL METHODS
Independent Variables: manipulated by
researchers
Dependent Variables: variables are those
assessed in response to the independent variables
Bina Nusantara University
5
RANDOM ASSIGNMENT & RANDOM
SELECTION
• Random assignment: occurs when people are
assigned to various treatment conditions or
levels of an independent variables in a
non-systematic way  every subject has an
equal chance of being assigned to every
condition.
• Random assignment is used as a means of
control by which groups of subjects can be
made more of less equivalent to one another
on variables not being studied.
Bina Nusantara University
6
• Random selection: we choose the subjects of
our investigation by a non-systematic method:
every possible subject of our study has an equal
chance of being chosen to participate.
• Random selection is important if we wish to draw
accurate conclusions about the entire group of
interest.
Bina Nusantara University
7
RESEARCH DESIGN
Research design: the basic structure of a
scientific study.
Experiment:
Survey Design:
Observation Design:
Bina Nusantara University
8
MEASUREMENT
• Measurement: the process of assigning
numbers to characteristics of people or things.
• Variables in every study must be measured or
quantified so that data analysis can be
conducted to draw conclusions.
• One of the most critical steps in planning a
research study is: DECIDING HOW each
variable will be MEASURED.
• Measurement can be classified as either
CATEGORICAL or CONTINUOUS
Bina Nusantara University
9
MEASUREMENT
CONTINUOUS measurement is used when the
numbers represent the amount of the characteristic in
questions. Higher number represent more of the
characteristic than lower number, so that inferences can
be made based on the value of a variable
 Are you hapy? (3) very happy – (2) happy – (1) not
happy
CATEGORICAL measurement: when the numbers DO
NOT represent an underlying characteristic than can be
measured continuously
 You received the information from: (1) books – (2)
friends – (3) media
Bina Nusantara University
10
MEASUREMENT
Reliability: the consistency of measurement
across repeated observations of a variable of
the same subject
As the error component increases, observations
will DIFFER each time the subject is assessed.
Internal Consistency Reliability:
Test-Retest Reliability:
Bina Nusantara University
11
Both internal & Test-retest reliability are
necessary properties for a useful measuring
device.
If a measure contains too much error, it will not
give sufficiently accurate measurement to be
useful
Reliability is NOT ENOUGH: just because a
measuring device is consistent does not mean
that it actually assesses the variable of interest
Bina Nusantara University
12
MEASUREMENT
Validity: represents the true score component
 the inferences made about a measuring device rather
thatn the device itself
 An intelligence test is considered valid if people who score
high do better than people who score low on tasks that in
theory require intelligence
Construct Validity: we are able to give an interpretation to
scores on a measure. To say that a measure has construct
validity is to say that we have CONFIDENCE in our
interpretation of what that measure represents.
Bina Nusantara University
13
Face Validity: means that a measure appears to
assess what it was designed to assess.
Content Validity: means that a multiple item
measure of a variable does an adequate of
covering the entire domain of the variable.
 a single question would generally be
inadequate to cover all the material in an entire
chapter of a text book.
Bina Nusantara University
14
MEASUREMENT
Construct Validity: we are able to give an interpretation to scores on a
measure. To say that a measure has construct validity is to say that we have
CONFIDENCE in our interpretation of what that measure represents.
Face Validity: means that a measure appears to assess what it was designed
to assess.
Bina Nusantara University
15
Content Validity: means that a multiple item measure of a
variable does an adequate of covering the entire domain of
the variable.
 a single question would generally be inadequate to cover
all the material in an entire chapter of a text book.
Criterion-related validity: means that score on a measure of
interest relate to other measures that they should relate to in
theory.
Face, Content, Criterion-related validity represents ways to
assess validity.
Construct validity is inferred based on research evidence. It
is our best guess about what a measure represents.
Bina Nusantara University
16
STATISTICS
Descriptive statistics summarize the results of a study.
* Measure of Central Tendency and Dispersion
* Correlation
* Regresssion
Inferential statistics help interpret the results using a variety of
statistical tests.
 Procedures that help you decide if the results can be
attributed to | error variance or the experimental treatment
If the probability of finding the mean difference by chance is
less than 1 in 20 (0.05), the conclusion is reached. 
Statistical Significance
Bina Nusantara University
17
5 INFERENTIAL STATISTICS TEST
(COMMONLY USED)
Independent Group t test
to determine if 2 groups of subjects differ significantly
on a dependent variable.
Analysis of Variance (ANOVA)
to determine if 2 or more groups of subjects differ
significantly on a dependent variable
Factorial ANOVA
to determine the significance of effects of 2 or more
independent variables on a dependent variable
Bina Nusantara University
18
t test for Correlation
to determine if the correlation between 2
variables is significantly greater than zero
Multiple Regression
to determine if 2 or more predictor variables can
significantly predict a criterion variable
Bina Nusantara University
19
META-ANALYSIS
A single study is never considered o offer a definitive
answer to a research question
To achiever confidence about a phenomenon of
interest, we need to conduct several studies.
A META-ANALYSIS: a quantitative way of combining
results of studies, much like our statistics summarize
the results across individual subjects
Bina Nusantara University
20
A Meta-Analysis can summarize statistically the
results of different studies in the domain of
interest to I/O Researcher. Such analysis can be
simple descriptive summaries of results or very
complex mathematical and statistical
procedures.
Bina Nusantara University
21
META-ANALYSIS
For example:
Suppose you found 5 studies that reported the
following correlation between job satisfaction and pay
level:.20, .22, .24, .26, .28
A simple meta-analysis of these 5 studies would
conclude that the mean correlation between these 2
variables was
.24
In most areas that have been frequently studied,
meta-analysis can be found to help interpret what
those individual studies have found.
Bina Nusantara University
22
ETHICS OF RESEARCH
The researcher must protect the well-being of
subjects.
 this means that manipulations, such as an
experimental training procedures, should be used if
they are known to cause harm
The researchers must take care to PROTECT
IDENTITIES when appropriate
It would be considered unethical to violate
confidentiality and disclose the identities of surveyed
employees.
Bina Nusantara University
23
The well-being of individuals against the wellbeing of organization has to be weighted
carefully
An ethical psychologist will discuss the issue
with other psychologists and with superiors in
the hope of reaching an ethical and satisfactory
decision.
Bina Nusantara University
24
ETHICS OF RESEARCH
It is good idea to try to foresee these situations and avoid
them. If you suspect that supervisors might demand to know
employee identities, conduct surveys anonymously. If you do
not know the identities, you cannot disclose them.
If there is even a slight possibility that there can be some
drawbacks to participation, the person should be asked to
sign an informed consent form.
These forms explain the nature of a study and what is
expected of the subject.
Bina Nusantara University
25