You are on page 1of 145

Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 5
Variables
Variables
 Aspect of a testing condition that can
change or take on different characteristics
with different conditions
 Types of Variables
 Dependent Variable (DV)
 Independent Variable (IV)
 Subject Variable
 Confounded Variables
 Extraneous Variables
Dependent Variable
 Measure of behavior that reflects the effects of
the independent variable.
Ways to Measure Behavior
 Frequency
 Rate
 Duration
 Latency
 Topography
 Force
 Locus
Independent Variable

 Conditions manipulated by the experimenter to


determine its effect.

 Levels
 Different values of the Independent Variable

 Confounded Variable
 Changes with the Independent Variable such
that their effects cannot be separated.
Describing Variables
 Quantitative – Varies in amount Vs.
 Categorical – Varies in kind

 Continuous – Not Limited in values Vs.


 Discrete – No Intermediate values

 Apparent Limits – One point (e.g., 4) Vs.


 Real Limits – Score +/- half the distance to the next
score (e.g., 3.5 – 4.5)
Types of Measurement Scales

Measurement is the process of assigning numbers


to things on a scale according to rules

 Nominal
 Kinds or types (red birds, blue birds)

 Ordinal
 Nominal + logical order (good, better)
 Interval
 Ordinal + numerical increments (thermometer)

 Ratio
 Interval + true zero (money)
Judging Measurements: Reliability
 Measurement gives the same result on
different occasions.
 Test-Retest Reliability
 Same test gives the same score on different
occasions
 Internal Consistency
 Each part of a test measures the same idea
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 6
Validity
Validity
 Degree to which research conclusions match
the real world

 Types of Validity
 Internal Validity (Handout )
 External Validity (Handout )
 Construct Validity
 Statistical Validity
Construct Validity
 Extent that the results support the theory behind the research
 Measures the idea it’s supposed to, and nothing else

 To improve construct validity


 Use a Manipulation check , which is designed to make sure that the
independent variable is changing in the way that it should
 Try to rule out other possible theoretical explanations of the result

 Construct Validity
 Face Validity
 Test seems superficially valid
 Content Validity
 Tests whole idea, not part
 Criterion Validity
 Test correlates to other tests of the same idea
Threats to Construct Validity
 Loose connection between theory and method.
 Poor operational definitions
 Poor measurement techniques

 Ambiguous effect of Independent Variable


 Can arise from the social aspect of an
experiment
Statistical Conclusion Validity

 Are the results due to chance or due to a cause and


effect relationship between the Independent Variable
and the Dependent Variable?

 Power is the ability to see an effect of the


Independent Variable on the Dependent Variable if it
is actually there.

 Effect Size is the strength of the relationship between


the Independent and Dependent Variable
Threats to
Statistical Conclusion Validity
 Erroneous choice of statistics
 If the wrong test is chosen, the results cannot
be trusted
 Limited Power
 The effect may not be observed because of too
few subjects
 Inaccurate Effect Size estimation
 The level of relationship between the
Independent and Dependent Variable is not
correct
Social Psychology of the Experiment
 Role Demands
 Good-Subject Tendency
 Role demands
 Evaluation apprehension

 Experimenter Bias – Experimenter


unintentionally biases the results. Control with
“blind” studies.
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 7
Control
Control
 Ways to rule out threats to validity

 Two basic meanings:


 Standard for comparison
 Something to compare the effects of the
Independent Variable to
 Control Group or Control Condition
 Minimizing variability
 Keeping things constant
Two Types of Experiments

 Within-Subjects Experiment
 Each subject experiences ALL experimental
conditions.
 Not always possible, due to contrast effects
or potential disturbances of subject naiveté

 Between-Subjects Experiment
 Each subject experiences only ONE
experimental condition
General Strategies to
Increase Control
 Use a lab, if possible
 Consider the preparation, or experimental setting
 Instrumentation for measuring responses
 Build Nuisance Variables into the experimental
design.
 Statistical control
 Replication
Specific Strategies for Control
in a Within-Subjects Design
 Each subject is used as their OWN control.

 Control Condition
 Occasion when the subject DOESN’T get the
Independent Variable.

 Experimental Condition
 The subject gets the Independent Variable.

 The Dependent Variable in the control condition is


compared to the Dependent Variable from the
experimental condition.
Specific Strategies for Control
in a Between-Subjects Design
 Experimental Group
 Subjects who get the Independent Variable

 Control Group
 Subjects who DON’T get the Independent Variable

 Random Assignment
 Every subject has an equal and independent chance to be in
a group.
 Matching
 Making sure that the groups are equal on one or more
variables before the experiment starts.
 Can weaken result if the matching variable isn’t correlated
with the dependent variable
The Rest of the Course
 The experiment as a
problem solving exercise
 Creating the elegant
experiment
 Choice of methods
 Review the literature
 Choice of subjects and/or
participants
 Ethical and practical
considerations
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 8
NonExperimental Research, Part 1:
Observational, Archival, and
Case Study Research
Non-Experimental Research
 Research in which the experimenter does not
manipulate the variables or assign subjects to
conditions.
 Also called qualitative research or correlational
research
 Hermeneutic Approach
 Asks what the behavior means rather than why
it happened.
Observational Research
 Watching and recording behavior, but not trying to change
it.
 Also called unobtrusive or nonreactive research

 Naturalistic Observation
 Subjects watched in natural environment
 Subjects disturbed as little as possible.

 Laboratory Observation
 Subjects watched in the laboratory

 Physical trace method examines evidence left behind by


behavior
Participant Observer Research
 Observer joins a group in order to record
behavior.
 Most useful when studied group is small or
remote
 Ethically controversial
 Invasion of participants’ privacy
 Potential experimenter objectivity difficulties
 Changing results simply by being a member of
the group.
Archival Research
 Attempting to answer research questions from
data that already exists - archival data
 Archival Data
 factual information in existing records or archives

Limitations of Archival Research


 Data collected for some other purpose
 Ruling out alternative hypotheses may be difficult
 No real possibility of informed consent
Case Studies

 Observation of a distinctive situation, usually


involving a single individual (person, animal,
institution, or group).
 Characterized by the uniqueness of the
individual.
 Narrative Case Study
 Story told first hand that reflects meaning
Theory Development
 Process is more flexible with non-experimental
designs but still involves hypothesis testing.
 Two approaches:
 Mathematical approach toward combating
threats to internal validity
 Inductive process of hypothesis generation and
modification
Recording Methods
Field Notes
 Be systematic
 Categories of behavior
 Length of observation period
 Checklists
 Be selective
 Take samples of behavior

 Make use of recording devices


Content Analysis
 Evaluation of a hypothesis using publicly
available pictures and language
 Manifest Content
 Measures the frequency of some word, image,
phrase, or action
 Latent Content
 Measures the appearance of themes, as
determined by the researcher
 Use at least two coders to increase reliablity
Suggested Guidelines for Research

 Maintain APA ethical standards


 Move from procedure to protocol
 Pilot the project before you begin
 Manage data carefully
 Lab notebook
 Care in data reduction
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 9
NonExperimental Research, Part 2:
Survey Research
Purposes of Surveys

 Determine how people feel about an issue


 Find out the effect of an event on behavior
 Examine correlations between responses
 Dispel myths
Designing a Questionnaire

 Determine the Purpose of the Questionnaire.


 What basic question is being asked?
 What variables might be of interest?

 Determine the Types of Questions


 Open-ended Question
 People can answer in their own words.
 Harder to analyze
 Closed-ended Question
 Easier to analyze responses
 Answers limited to the choices available.
Write out the Questionnaire Items
 One issue per item
 Avoid bias
 Make closed-ended question alternatives clear
 Mutually exclusive
 Exhaustive
Beware of…
 Social Desirability
 One answer is more socially acceptable than
others
 Counter with a Verification Key
 Also called a lie scale
 Collection of items that detect dishonest answers

 Acquiescence
 Tendency to agree with any statement
Determine the Format of the Item
 Define key terms
 Closed-ended formats
 Visual Analogue Scale

 Likert Scales

 Branching Items

 Open-ended formats require space to write


 Sequence the all items - Put things in the right order
Example
Determine How the Data Will Be
Analyzed
• Determine Scoring
• How will each question be represented in for
analysis?
• Use a Coding Guide to write it down!

• Data Reduction
• Take care moving from individual sheets to
analysis software
• Determine Statistics
• What will you analyze and how?
Administering the Questionnaire
 Methods of Survey Administration
 Face to Face
 Written Responses
 Computerized
 Telephone

 Response Rate
 The percentage of individuals who return the
completed survey.
 Prefer at least 50% for research purposes
Face-to-Face
 Advantages
 Interviewer can
 Establish rapport with people
 Direct attention to material
 Guarantee the order of questions
 Disadvantages
 Interviewer can create a social situation that
leads to response bias
 Expensive
 Difficult to supervise
Written Responses
 Advantage
 Relatively cheap, even if mailed
 Anonymous, so response bias reduced

 Disadvantage
 Low response rate; often less than 50%
 Cannot clarify misunderstood questions
 Cannot tell how seriously the person took the
questions.
Computerized Administration
 Advantages
 Impersonal, so social desirability reduced
 Order of completion guaranteed
 Internet versions available 24/7
 Inexpensive

 Disadvantages
 People may not understand questions
 Response rates are low
 Random sample difficult to obtain
Telephone Administration
 Advantages
 Questions can be clarified
 Can be performed with computer assitance
 Supervision of interviewers possible

 Disadvantages
 Survey must be short
 Cell phone minutes cost
 Surveys less anonymous
People & Surveys
 Population
 All of the people or animals in the group involved
in the research
 Sample
 Sub-group of the population.
 May be selected in several different ways:
 Haphazard sample
 Purpositive sample
 Convenience sample
 Probability sample
Probability Samples
 The probability of any element being selected
in a sample is known.

 Several types:
 Systematic Sample
 Simple Random Sample
 Stratified Random Sample
 Cluster Sample
Random Samples
 Simple Random Sample
 Each member of the population
has the same chance to be
chosen, independently of any other
member
 Stratified Random Sample
 A special type of random sample in
which at least 2 subgroups are
represented in the same
proportions as the population.
Non-Random Probability Samples
 Systematic sample
 Used for practical reasons
 Likelihood of any element being selected is still
known
 Cluster sample
 Groupings from a larger population
 Multistage sampling
 Clusters further broken down by taking samples
from each cluster
Obtaining True Answers to Questions

 Randomized Response Method


 A way to estimate the response level in the
sample without knowing the response of any
particular individual
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 10
True Experiments, Part 1:
Single-Factor Designs
True Experiments

 Scientist has complete control over all


aspects of the experiment.
 Subjects are randomly assigned to conditions
 Researcher has control over what, when,
where, and how of the experiment
 Different from Quasi Experiments (Chapter 13)
where control less
Independent Variables
 Factors
 Independent Variables.
 Values of factors are called levels
 So a Single-Factor Design means only one
Independent Variabke.

 Condition or Treatment
 Way that the subjects experience the factor
 Experimental or Control
Elements of True Experimental Design
Within-Subjects Design
 Important to control for possible carryover
effects (order and sequence) through
counterbalancing.
 Order Effects
 the ordinal position of a condition changes a
subject’s performance
 Sequence Effects
 conditions interfering with performance on
each other.
Controlling Order and Sequence Effects Within Subjects

 Randomization of the order of conditions

 Block Randomization
 Order of conditions is randomized, but each
subject is presented with each condition before
any condition is repeated.
 Example: BCAD, ADCB
Reverse Counterbalancing
 Way of controlling carryover effects within
subjects
 Conditions are presented first in one order, then in
the reverse.
 Example: ABCCBA
 Works well with potential linear confounders, but
not practice effects.
Controlling Order and Sequence Effects
Between Subjects
 Counterbalancing Completely
 Each condition occurs an equal number of times
in each rank-order position and follows each
other condition an equal number of times.
Controlling Order and Sequence Effects
Within GROUPS
 Latin Square
 Incomplete counterbalancing in which each
condition is experienced by one subject in a
systematically different order from other
subjects.
Two-Conditions,
Within-Subjects Design
 Simplest true-experimental design
 Subjects experience all conditions
 Two conditions:
 Experimental
 Control

 Conditions counterbalanced across subjects


 Can be modified for multiple conditions
 Several levels of the independent variable, as well as a
control condition.
Two Conditions
Within Subjects

 Reverse
Counterbalancing
 Picture, Imagine,
Imagine, Picture
Two-Conditions,
Between Subjects
 Each subject in only one group

 Subjects randomly allocated to groups

 Two groups:
 Experimental
 Control

 Can be modified for Multiple Groups


 Several levels of the independent variable as
well as a control group.
Multiple Conditions,
Between Subjects
 Electro
convulsive
shock
 Number of
explorations
as a function
of interval
between ECS
and learning
Designs to Avoid
 One-Group Posttest-Only Design
 Measures behavior of a single group of
subjects after they’re given a treatment
 Nonequivalent Control Group Design
 Group of subjects that are not randomly
selected from the same population as the
experimental group
 One-Group Pretest-Posttest Design
 Measures behavior of a single group before
and after treatment
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 11
True Experiments, Part 2:
Factorial Designs
Factorial Designs
 Designs with more than one factor
(independent variable)
 Each factor is represented by a number that
refers to the number of levels
 e.g., 2 x 2 x 2 factorial design
Factorial Designs

 Main Effects
 Effect of one factor, averaged over the effects
of everything else.
Interactions
 Effect of one IV depends on the effect of
another IV
Interactions and Main Effects
 Possible to have one without the other, so in a
2 x 2 experiment, eight outcomes are possible
Example 2 x 2 Experiment

Outcome of Example Experiment.


(a)Relationship Between Attraction Within
Couples and Distance, with True Love as a
Parameter.
(b)Relationship Between Attraction and
True Love, with Distance as a Parameter.
Antagonistic Interactions
Type of Interaction when 2 IVs tend to
reverse each other’s effects
Synergistic Interactions

Type of Interaction in which 2 Independent


Variables tend to increase each other’s effects

Graphs of a Synergistic Interaction.


(a) Plotted with Response as a Function of A, with B as a Parameter.
(b) Plotted with Response as a Function of B, with A as a Parameter.
Ceiling Effect Interactions

 Type of Interaction in which one Independent


Variable has a smaller effect when paired with
a second variable

Graph of a Ceiling-effect Interaction.


(a) Plotted with Response as a Function of A, with B as a Parameter.
(b) Plotted with Response as a Function of B, with A as a Parameter.
Within Subjects Factorial Designs

All subjects get all


conditions

Possible Sequence of
Conditions Experienced by
Eight Subjects in the Factorial
Experiment Illustrated Above
Simplified Design of the
Stevens and Rubin Experiment
Between Subjects Factorial Design

 Each subject gets


one condition

Design of the Wansink and Kim (2005) Study


Mixed Factorial Design
 Subjects get at
least one within-
subjects factor and
one between-
subjects factor

Relative Time to Apply Brakes in Light


Traffic and Heavy Traffic While Talking
on a Cell Phone (Dual Task) or Not
(Single Task)
Considering the Number of Subjects in
Factorial Designs
 Within-Subjects requires the fewest number of
subjects

 Between-Subjects requires the highest number


of subjects

 Mixed-Designs require less than Between-


Subjects, but more than Within-Subjects
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 12
Single-Subject Designs
Advantages
to Single-Subject Designs
 Doesn’t distort the behavior, which can happen when
examining group means

Hypothetical Learning Data for a Group Individual Participants in


Group

 Effects must be large enough to reach clinical


significance.
 Practical and Ethical Problems can be addressed.
 Flexibility in Design.
Disadvantages to
Single-Subject Designs

 Sometimes, a subtle effect is still important.


 Some questions require groups to answer.
Baseline Measures
 Control strategy
 Two purposes:
 Descriptive
 Predictive

 Should be stable before beginning experiment or show


a trend that is opposite to the predictive baseline.
 Trend
 Variability that is systematic.
 Indicates a distinctive direction in the DV.
 Can be Ascending or Descending
Intervention
 Repeated measures of behavior (Dependent Variable)
 Behavior measured under a treatment (Independent
Variable)

 Should cause a change from the descriptive baseline in


the direction of the predicted baseline
AB Design
 Also called Comparison Design
 Baseline followed by Treatment
 You cannot tell if the changes in the dependent
variable were due to the independent variable
ABA Design
 Baseline, Treatment, Removal of Treatment

 Problems with this Design


 Treatment may not be reversible (e.g., learning)
 You may want to leave the learner in the new
condition
ABAB Design
 Also called a Repeated Treatments Design or a
Replication Design
 Baseline, Treatment, Removal of Treatment, Treatment
Repeated
 Leaves the learner with the benefit of treatment
Alternating Treatments Design
 Allows the comparison of two different
Treatments (or Independent Variables)
Alternating Treatments Design
 Employs multiple treatments without violating the “only
change one variable at a time” rule
 Baseline, Treatment 1, Treatment 2, Treatment 1,
Treatment 2, etc.
 Used to assess which treatment is most effective
Multiple Baseline Design
 Tries to see if behavior
change corresponds to
onset of treatment
 Multiple Baseline Across
Behaviors
 Different behaviors
 Multiple Baseline Across
Individuals
 Different subjects
 Multiple Baseline Across
Settings
 Different locations
Changing Criterion Design

 Often used in shaping behavior

 Introduces successively more stringent criteria


for reinforcement to see if behavior changes
with the criteria
 Useful when behavior change is irreversible,
as in learning
Visual Threshold Tests
 Humans
 Staircase technique
 Factorial Single-Subject
Design
 Pigeons
 Blough (1956) trained
birds to peck in one place
until they couldn’t see the
light, then to switch to a
second place.
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 13
Quasi Experiments
Quasi-Experiments
 Research in which the scientist does not have complete
control over the who, what, where, when, or how involved in
the study.

 Lack at least one of a true experiment’s defining


characteristics.

 Subjects selected rather than randomly assigned to


conditions

You can’t be 100% sure that the changes in the Dependent


Variable are due to the Independent Variable with Quasi-
Experiments.
Nonequivalent Control Groups

 Most common type of quasi-experimental design


 Has an Experimental group
 Has a Control group
 Subjects NOT randomly assigned

How do we know that the subjects were


equivalent (as far as the IV goes) before the
treatment? We don’t!
Patterns of Nonequivalent Control Group
Design with Pre-test/Post-test
Interpretable

Uninterpretable
Mixed Factorial Design with
One Nonmanipulated Variable
 Type of nonequivalent
control group quasi-
experimental design.
 More than one
independent variable.
 Subjects NOT randomly
assigned for one factor.
 Other factors are true
experimental variables.
Interrupted Time Series Design

 No control group
 Same group compared over time before and
after a manipulation
 Baseline measurement, Treatment
Repeated-Treatment Design
 Baseline
 Treatment given
 Withdrawn
 Treatment given again

 No control group
Cross-Sectional Studies
 Developmental Design Percentage of People who
 Study several groups of can Program a DVR
people at different ages 2000 2010 2020
all at the same time. 1960 10 40 70
 Possibly confounded with 1970 40 60 80
Cohort Effects.
1980 70 80 90
 A cohort is a group that
has something in common
1990 100 100 100
Longitudinal Studies
 Developmental Design Percentage of People who
 Study the same group of can Program a DVR
people over time. 2000 2010 2020
 Possibly confounded with 1960 10 40 70
Secular Trends.
1970 40 60 80
 A secular trend is a change
that is taking place in the 1980 70 80 90
general population over
time 1990 100 100 100
Cross-Sequential Design
 Developmental Design Percentage of People who can
 Test 2 or more cohorts at 2 or Program a DVR
more times (same ages) 2000 2010 2020
 Design used to help 1960 10 40 70
separate developmental,
1970 40 60 80
cohort, and secular effects.
 Time Lag Effects are the result of 1980 70 80 90
comparing subjects of the same
age at different times 1990 100 100 100
Program Evaluation

 Techniques to determine the effectiveness of a


social service program.

 Often hindered by politics within the agency.

 People who stand to gain or lose through an


evaluation of the program are called
Stakeholders.
Program Evaluation

Techniques for determining the effectiveness of a social


service program
Steps in Executing
1. Identify Stakeholders.
2. Arrange preliminary meetings with Stakeholders.
Who wants the evaluation? Why?
3. Decide whether to do the Evaluation.
4. Examine the literature.
5. Determine the methodology.
6. Present a written proposal.
Meta-Analysis
 Set of methods for
combining the results
of many different
studies
 Examines a larger
view of the topic
 Effect size vs
significance
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 14
Data Exploration Part 1: Graphic and
Descriptive Techniques
Preparing Data for Analysis
 Design the study BEFORE you start it – including
statistics!
 Write it all down in a lab notebook
 Once the data is collected:
 Put the data into a summary data sheet.
 Do preliminary statistics and plots.
 Check for invalid data
 Check for missing data
 Check for wild data
 Describe data numerically.
 Describe data graphically.
 Perform inferential statistics.
Data Reduction
 Process of transcribing data from individual
data sheets to a summary form or data file.
Coding Guide
 List that specifies the variables of the study,
columns that the variables occupy in the data
file, and their possible values.
 Located either on summary form, in notebook,
or both!
Measures of Central Tendency

Descriptive statistic that is the average of the


distribution.

 Mode = Most common score

 Median = Middlemost score

 Mean = Sum of all the scores divided by


the number of scores.
Measures of Variability
 Range
 Highest score – Lowest score

 Percentile
 Score below which a certain number of cases in a
distribution fall
 Interquartile Range
 75th percentile – 25th percentile
 Q3 – Q1

 Semi-interquartile range
 (Q3 – Q1)/2
Most Common Measures of Variability

 Variance
 Average of the squared deviations from the mean.

 Standard Deviation
 Square root of the variance.
Tables

Displays of data in matrix format


Graphs

Data in spatial relationships in a diagram


Frequency Table
Frequency Distribution
 Graph that shows how many scores fall into
particular bins, or divisions of the variable

Histogram
Frequency Distribution

Polygon
The Shape of Distributions

Normal Negative Positive


Skew Skew
Cumulative Frequency Distribution

a = Normal Curve
b = Positively Skewed
c = Negatively Skewed
Scattergram
 Often used with correlation coefficient
 A correlation is a statistic indicating the strength
of the relationship between two variables
 Prediction of one of the variables can be
achieved with regression
Correlation Coefficient
 Measures strength of association between variables.
 Does NOT indicate causation
 Most commonly used is Pearson’s r
 Value is between – 1.0 and +1.0

Scattergram of Paired Values of x and y; (a) r = +1.00, (b) r=−1.00, (c) r = 0.50, (d) r = 0, and (e) r = 0
Regression
 Predicting the value of one variable from
another from the equation for a line

 Slope of the line (m) reflects


 Correlation
 Scale of measurement for the two variables

 Squaring the correlation yields a goodness of


fit measure
Line Graph
 Y-axis is the dependent variable
 X-axis is the independent variable
Bar Graph
 Used to represent categorical data
Frequency Data and Graphs

Test Score as a Function of Class Membership

Frequency Distribution of Test Scores by


Class Membership
Time Series Graph
 X-axis represents the passage of time

Time-Series Graph Cumulative Record


Indicating Variability
 Error Bars
 Vertical lines above and below each point or
bar on a graph that show +/- one standard
deviation from the mean.
 Can be misleading
Box and Whisker Plot
 Graph based on median and percentiles rather than mean
and standard deviation.
Checking for Problem Data

 Invalid Data
 Outside the range of possible values
 Find and correct

 Missing Data
 Empty cells
 If necessary, replace with code
 Outliers
 Possible, but improbable answers
 Check to see if they are different enough to
remove
Style Guide for Figures
 Be clear
 Use black ink
 Label both axes
 Label units of measurement
 Provide a caption for the figure
 Beware of chartjunk (parts that aren’t
necessary to understand the chart)
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Chapter 15
Data Exploration Part 2:
Inferential Statistics
Statistics
 Quantity computed from a sample

 Descriptive Statistics
 Summarize a set of data
 Chapter 14

 Inferential Statistics
 Assist in drawing conclusions about populations
Sampling Distributions
 Distribution of
means of samples
from a population
 Standard Error of the
Mean
 Standard Deviation
of a sampling
distribution
Hypothesis Testing

 H1 (Alternative) vs H0 (Null)
 Statistical significance is the probability that a result happened
by chance
 Alpha is the probability of deciding that the null hypothesis is
false when it’s actually true.
Significance
 Probability that an experimental result
happened by chance
 Generally alpha less than .05
 Does not necessarily mean that the result was
important or large
 Size of result is measured with Effect Size
 Effect size shows the strength of the
relationship between the independent and
dependent variables.
Power of the Test of the Null Hypothesis
Against the Alternative
 Power is the probability of rejecting the null
hypothesis when it actually IS false
Chi Square Statistic
 Tests Frequency
Data to determine
whether two
categorical
variables are
related.

 Expected
Frequency vs
Observed
Frequency
ANOVA
 Compares more than two conditions
 For only two, use a t-test

 Tests the significance of a difference among


several conditions in an experiment by making
two different estimates of the variability that
would be expected if the null hypothesis is
true.
 Between variability
 Within variability
How to Read an ANOVA Table
Between Subjects Single Factor Design
How to Read an ANOVA Table
Within Subjects Single Factor Design
How to Read an ANOVA Table
Between Subjects Factorial Design
Example
 Motor co-
ordination
measured in
groups of people
who all
experienced
different dosages
of marijuana
Research Methods, 9th Edition

Theresa L. White and Donald H. McBurney

Epilogue
Biases and Limitations of Experimental
Psychology
Science as Conservative
 Scientists can be
 Older
 Unwilling to change their theoretical positions
 Research funded by the Federal Government
 Subject to political pressure
 Potential for Bias
 Some Research Funded Corporately
 Potential for Bias
Science as Liberal
 Force for Change in Establishment
 Sometimes more radical than necessary
 Treating conformity as a weakness
Limitations of Science
 Essential Limitations
 Neutral in the question of values
 Culturally relative
 Tentative

 Practical Limitations
 Opportunistic nature
 Cost of research
 Complexity of problem
Responsibilities of the Scientist

 Society must be benefited by the work

 Free speech must be used

 Educate the public of your findings

You might also like