Research Methods Exam 3

Front 1

Quasi-experiments

Back 1

Experimenter doesn't directly manipulate variables

Variable are selected

Front 2

Natural Treatment Design

Back 2

Design: observation to treatment to observation

E.G. effect of new curriculum on 3rd grade achievement

Note: this is not a true reversal (ABA) design

Front 3

Natural Treatment

Problems (potential confounds)

Back 3

Maturation-changes in the participants themselves

(e.g., experience with tests, better social adjustment, biological maturation)

History-influences outside the classroom that could affect results

Front 4

Natural Treatment

Solutions

Back 4

Use nonequivalent control group

But

No random assignment

Matching after the fact; therefore, selection bias

Front 5

One-shot case studies

Design and Problems

Back 5

Design: History to current behavior

E.G. Clinical Case

Problem: No baseline observations

Thus, no control condition

Front 6

One-shot case studies

Solutions

Back 6

Use Deviant Case Analysis

Similar individual without treatment, as control

Need many observations and good detective work

e.g., suppose patient is depressed; history shows

• Abusive parent
• Family history of depression
• low socioeconomic status
• Was a passenger on a hijacked airliner
• Exposed to toxic chemicals on the job site
• Family moved often

Front 7

Interrupted-Time Series Design

Back 7

Does Flouride cause increased achievement?

Third variable problem (e.g., absenteeism)

Use nonequivalent control group

Examine several dependent variables

Other sources of confounding:

Other changes that occurred

Mortality (not same subjects before and after the tests)

Front 8

Designs Employing Subject Variables

Back 8

Subject Variable

Some measurable characteristic of people

(I.Q., sex, age, race, addiction, mental illness)

Always confounded with other variables

i.e., can't manipulate subject variable while holding other factors constant

Use matching to control for confounded variables

e.g., compare schizophrenics to neurotics of similar age, I.Q., length of hospitalization etc.

Front 9

Designs Employing Subject Variables

Problems with matching

Back 9

Relevant variables may be unknown

May result in small n's

Subtle interactions among matching variables may confound results

Regression artifacts

Front 10

Designs Employing Subject Variables

Age as a variable

Back 10

Cross sectional Design

e.g., compare 5,8, and 12 year olds

Problem: age confounded with generation of birth

(different attitudes, education, etc.)

Front 11

Designs Employing Subject Variables

Longitudinal Design

Back 11

Same subjects studied over time

Problem: age confounded with time of testing

Front 12

Designs Employing Subject Variables

Time lag Design

Back 12

Examines time of testing while holding age constant

Problem: confounding time of testing with generation gap

Front 13

Designs Employing Subject Variables

Cross-sequential Design

Back 13

Test 2 or more age groups at 2 or more times

Includes features of other three designs

Different ages tested at the same time (cf X- sect.)

Same subjects tested successively (cf longitudinal)

Different subjects of same age tested at different times (cf time lag)

Therefore, can determine effects of major confoundings

Front 14

Designs Employing Subject Variables

Alternative

Back 14

Cross Sectional design with age as quasi-independent variable plus another true independent variable

Look for interaction: e.g., effect of two learning strategies and age on memory

If interaction between strategy and age, try to determine what variable confounded with age is responsible

Front 15

Designs Employing Subject Variables

Protocol 1

Back 15

Age

5 yr Memory Test

12 yr Memory Test

Results: 12 year old > 5 year old

Interpretation:

Brain Maturation?

Education?

Reading Level?

Motivation?

Experience?

etc.

Front 16

Designs Employing Subject Variables

Protocol 2

Back 16

Age: Training:

5 yr. Rote vs. Categorization vs None

12 yr. Rote vs. Categorization vs None

Memory Test

Results:

Rote: 5 year old > 12 year old

Categ.: 12 year old > 5 year old

None: 12 year old > 5 year old

What variable associated with age accounts for interaction?

Front 17

Operant Conditioning

Basic Paradigm

Back 17

Basic Paradigm: R to SR+ to increasing probability of R

Importance of controlled environment (skinner box)

Focus on rate of response

Front 18

Operant Conditioning

Importance of schedule of reinforcement

Back 18

Strongly influences temporal pattern of response

Generality established

Across different species

Across different responses

Across different reinforcers

Front 19

Operant Conditioning

Focus on behavior of individual subjects

Back 19

Reliability demonstrated by stabililty of baseline

Across sessions

Across individuals

Group comparisons not used

No group means

No tests of statistical significance

Front 20

Operant Conditioning

Applications

Back 20

Behavior modification

Programmed learning

Behavioral pharmacology

Front 21

Operant Conditioning

The Reversal Design (ABA)

Back 21

(A) Baseline

(B) Independent Variable

(A) Baseline

Front 22

Operant Conditioning

Multiple-Baseline Design

Back 22

If permanent carry over effects

Front 23

Interpreting Results

Back 23

Interpreting specific results

Scale attenuation effects

Types

A) Ceiling effects

B) Floor effects

Causes

a) scale of measurement not broad enough

b) task difficulty (too easy or too hard)

Take home message- Always ask: are the results limited by the scale of measurement of the task?

'0' Retention level- possible ceiling effect

9-18 level- floor effect?

Solution- run pilot study and adjust scale or task accordingly

Front 24

Interpreting Results

Regression Artifacts

Back 24

1. Statistical regression to the mean
   
2. Likely to occur in quasi-experimental designs where subjects are matched
   
3. Solution- random assignment of subjects to groups
   

Front 25

Interpreting Patterns of Research

Back 25

Reliability and Replication

Reliability increased when

a) large number of observations

b) repeatable results

Replication results=experimental reliability vs. statistical reliability

Three kinds of replication

a) direct-repeat experiment as closely as possible

b) systematic- repeat experiment while varying factors considered irrelevant

c) conceptual-replicate the phenomenon in an entirely different way

Front 26

Interpreting Patterns of Research

Converging Operations

Back 26

1. Definition- a set of independent operations that eliminate alternative explanations of experimental results
2. Purpose- to validate an explanatory concept

a) e.g., Stroop effect: Increased time to name a color with the ink and color names don't match

b) Interpretations:

i) Perceptual inhibition- reading inhibits perception

ii) Response competition-between two different names that are elicited

Front 27

Which small-n-design is most susceptible to confounding from the order of conditions?

Back 27

AB

Front 28

In an ABA design, we can conclude that it was the treatment variable that affected the change during phase B if

Back 28

behavior in the third phase did return to the baseline

Front 29

Which of the following is generally agreed to be a characteristic disadvantage of (or threat to validity in) a reversal design

Back 29

The possibility of carryover effects

Front 30

Which statement is true of within-subjects multiple baseline experiments?

Back 30

Effects cannot be attriubuted to the independent variable if the measured behaviors interact.

Front 31

Which of the following is characteristic of changing-criterion design?

Back 31

The researcher manipulates the contingency for a particular outcome.

Front 32

A quasi-experiment differs from a true experiment because in a quasi-experiment

Back 32

Researcher selects levels of the independent variable

Front 33

In an interrupted time series design, there are all of the following except

Back 33

Interruptions caused by uncontrolled variable other than the treatment

Front 34

The phenomenon of regression to the mean implies that if your parents are slightly more intelligent than the average, your intelligence is probably:

Back 34

above average but less than your parents'

Front 35

You are interested in surveying attitudes about gambling as a function of age. Assume that you decide to conduct a poll today about gambling expenditures. This poll will consist of an equal number of people at 18, 24, 30, and 60 years of age. This is an example of which age-related method?

Back 35

cross-sectional

Front 36

An interaction between age and an experimental treatment means that:

Back 36

The effect of the experimental treatment differed at various ages

Front 37

The issue of scale attenuation in the dependent variable refers to the problems of interpretation when performance is:

Back 37

either nearly perfect or very poor

Front 38

Suppose that you were an A student, but on your last test you received a B, and the average of the class was a C. The phenomenon of statistical regression suggests that on your next test you will most likely receive a(n)

Back 38

A

Front 39

Experimental reliability may be established by:

Back 39

replication with an additional sample of similar participants

Front 40

Systematic replication of an experiment involves:

Back 40

changing numerous variables not thought to be critical to the phenomenon under consideration

Front 41

Converging operations are important in psychology because:

Back 41

they serve to eliminate alternative explanations of a phenomenon