Final Flashcards
1
Q
Matching Techniques
A
- Holding Variables Constant
(i. e. hold gender constant by testing females only) - Build Extraneous Variables into research theme
(i. e. test both genders)
- create a second (non-manipulated) IV
- “statistical control” for some quantitative variables - Yoked Control Procedure
- each control subject “yoked” to an experimental subject - Equating Participants (Best, if possible)
- matches subjects on the variable to be controlled
2
Q
Disadvantages of Equating Participants Technique
A
- Difficult to determine which variables are most critical
- Difficulty of finding “matched” participants increases as number of matching variables increases
- Such matching decreases generalizability
- Some variables very difficult to match participants on
3
Q
Counter balancing
A
- Randomized
- Complete
- Incomplete
- Intra-subject
4
Q
One Group Designs
A
-weak
- Posttest-only Design
- Pretest-Posttest Design
- Nonequivalent Comparison Group Design
5
Q
Postest-only Design
A
X O
X is the treatment or the IV
O is the observation of changes in the DV
No control condition to know if the IV affects DV or if effects were caused by some other variable
6
Q
Pretest-Postest Design
A
O X O
Observed change in DV from pretest to posttest to see if IV has an effect on DV
Could be several alternative explanations
-Products of “threats to internal validity”
7
Q
Nonequivalent Comparison Group Design
A
-common
Posttest only Pretest-Postest test
Experiment X O O X O
Comparison O O O
- not RA
- often self-selected groups
- problem of selection differences
8
Q
Experimental Group Designs
A
-strong
- Within-Participants Designs
- Factorial Designs
- Between-Participants Control Group Designs
9
Q
Within-Participants Designs
A
- Characteristics
- “Repeated measures” over time
- All subjects receive all conditions
- Fewer subjects
Disadvantages:
- taxing on subjects
- sequencing effect problems
10
Q
Factorial Designs
A
2+ IVs, at least 1 is more manipulated
11
Q
Between-Participants Control Group Designs
A
Posttest only Pretest-Postest test
Experiment X O O X O
Control O O O
- RA
- Advantages and disadvantages of pretest
12
Q
Survey Questions
A
- open-ended
- -hard to score
- closed-ended
- yes/no
- ranking
- rating
- Likert/intensity scale formats
- semantic differential format
- -better to have no ticks in the middle
- -ex. |————————————–|
13
Q
What is used to evaluate behavioral items of a survey?
A
Frequency
14
Q
Surveys Parts
A
- title and seal
- appeal and instructions
- -short and easy as possible
- headings and subheadings
- transitions
- -few
- response directions
- bold typeface
- justification of response spaces
- shading
- white space
- printing
- font type
Look for normal distribution of answers to determine validity of questions
- differential weighting
- -reported via histogram
15
Q
Pretesting surveys purposes
A
- Identify sources of error
- Examine effectiveness of revisions
- changes should be piloted with a different group (naive subjects) - Indicate the effect of alternate versions
- Assess the final version of a questionnaire for respondent understanding, time, and ease of completion
- . Allow the survey analysts to make changes to the format that might make data entry or analysis more efficient
16
Q
Methods for pretesting surveys
A
- One-on-one interviews
- Respondent focus groups
- Behavior coding respondent-interviewer interactions
- Interviewer and respondent debriefings
- Item non-response rate
- question skipped a lot? less valid
17
Q
Administering the survey
A
- Selecting appropriate respondents
- is the sample representative of the population?
- sample should OVERSTATE the diversity - Choosing a sample size
- Distribute the survey
- standardization
Probability sampling -simple -systematic (ex. exit polling) -stratified -cluster Non-probability sampling -convenience --often used with probability sampling
18
Q
Coding, Entering, and Analyzing Survey Data
A
Process includes:
- coding the survey questions
- entering data into spreadsheet
- -2 people
- verifying data are entered correctly
- -random checks of a sample of questions
- conducting statistic analysis
- -t-test
- ANOVA
- interpreting the results
19
Q
Trend/marginal effect
A
Called this if not statistically significant but is between 0.05 and 0.10.
20
Q
Statistical Analysis and Interpretation
A
- response rate
- reliability
- sampling error
- uni-variate analyses
- -frequency counts, means, modes, medians
- bi-variate and multivariate analyses
- -ANOVA
- -i.e. correlation coefficients, chi-square, t-tests
21
Q
Presenting surveys
A
Reporting results to:
- those who commissioned the survey (told at the beginning)
- academic journals
- conferences, colloquiums
- the public at large
22
Q
Survey Reliability and Validity
A
Survey reliability
- consistent data
- 2 types of error:
1. Random error - unpredictable
2. Measurement error - associated with how a survey performs in a particular population - -highest in construct validity
Survey validity
- accurate validity
1. Face - pilot subject
2. Content - in survey
3. Criterion-related - predictive (ex. retest in future)
4. Construct
23
Q
The Survey Research Method
A
- Specific and measurable
- Straightforward
- Pretested to ensure no unclear questions or incorrect skip patterns
- Administered to adequate population or sample of respondents so generalizations can be made
- Appropriate analysis to obtain objectives
- Accurate reporting of results (verbal and results)
- commissioners - Reliable and Valid
24
Q
Simple-Subject Designs
A
- statistics not required (rare)
- one subject or small group of subjects
- “Time-series” designs
- data analyzed via visual inspection
- methods of experimental control involve how the IV is exposed and withdrawn over time
- comparison = subject’s pre-treatment responses
- Baseline (the alpha condition)
- ABA and ABAB designs
25
AB Design
Single introduction of at least one IV on one baseline
Poor because of alternative explanations
26
ABA Design
Intro and subsequent removal (withdrawal) of an IV
- reversal is the diminishing of the second A
- irreversibility is the maintaining of the second A
Problems:
- ends on baseline condition (therapeutic issue)
- failure to return to baseline due to carryover effects (i.e. gambling)
27
ABAB Design
Best
Usually animals or children on the autism spectrum
28
Developmental Research Designs
1. ABAB Interaction Design
- combined effect of multiple IVs
- each IV presented separately and together
2. Changing Criterion Designs
- changing requirements on subjects for reinforcements
- Initial baseline
- Performance criteria change over time
3. Multiple-Baseline Design (most common)
- sequential introduction of an IV across more than one baseline
- multiple subjects, behaviors, or situations
a) same behavior: 2+ individuals
b) 2+ different behaviors for the same individual
c) same behavior, same individuals, different situations
29
When is extinction used?
When you can't use punishment
30
Instinctive Drift
Return to instinctive behavior
31
Techniques to make sense of raw scores
Number assigned to overall performance on a test
- Frequency Distributions - orderly arrangement of group of scores
- Descriptive Statistics - describe or summarize a distribution of test scores numerically
- The Normal Curve - theoretically distributions that are perfect and symmetrical
Can calculate descriptive statistics
- central tendency: average scores
- variability: spread of scores
- relationship
Can compare to norms
32
Graphic Representation of Data
```
Bar graphs (categorical variables)
Histograms (quantitative variables)
Line graphs (quantitative variables)
-dimension over time
-interaction between 2 variables
Scatterplots (2 quantitative variables)
```
33
Scatterplots for correlational data
Correlation coefficient - statistic to describe relationship
positive correlation: /
negative correlation: \
34
Bar graph
Restate hypothesis before bar graph
35
Procedures for interpreting scores
Normal probability distributions are theoretical distributions (hypothetical)
- perfect and symmetrical
- bell-shaped curve
- -asymptotes
- tests yield real scores from real people
36
Central Tendency
The shape of a score distributions affects the relationship between mean, median, and mode
37
Measures of variability
Variability - graph width
Range
Variance
- large variance = scores farther from the mean
- affected by outliers
```
Standard Deviation (sqrrt[var])
-how far, on average, each score varies from the mean
```
Curves less reliable in tails
38
Characteristics
"clinical cases" = 2.5% (tails)
"typical" = body
curve is convex at the highest point and becomes concave at 1 standard deviation above/below the meean
asymptotes
39
Score %s on a normal distribution
50% of scores above the mean, 50% below
34. 1% of population will score between the mean and one standard deviation above the mean
13. 6% between 1 SD and 2 SD
2. 1% between 2 SD and 3 SD
0. 8% between 3 SD and infinity
