Midterm Flashcards
Belmont Report
1) Beneficence: risk-benefit analysis of findings vs. harm
2) Autonomy: respect for participants and their decisions
3) Justice: fairness in accepting risk and receiving benefits
APA Code of Ethics
1) Beneficence: risk-benefit analysis of findings vs. harm
2) Fidelity and responsibility: maintaining trust and following through
3) Integrity: don’t lie, cheat, plagiarize, etc.
4) Justice: fairness in accepting risk and receiving benefits
5) Respect: respecting individual differences, respecting consent, being aware of own biases
Six steps of a research project
1) Ask a question stemming from a theory
2) Develop a specific and testable hypothesis
3) Select a method and design the study
4) Collect the data
5) Analyze data and draw conclusions
6) Report findings
How do we minimize harm?
1) Informed consent
2) Debriefing
3) IRB
What defines experimental design?
Must have manipulation of independent variables and random assignment
What is a quasi-experimental or subject variable?
A trait that cannot be changed about the participant, but participants can be grouped based on these traits (height, shoe size, age, eye color, etc.)
Internal validity
The extent to which causal conclusions can be substantiated
External validity
The extent to which results can be generalized
Construct validity
The degree to which variable operations accurately reflect the construct they’re designed to measure (free from systematic error)
Criteria for causality
1) Relationship between variables
2) Causal variable precedes affected variable
3) No possibility of a third variable affecting both (confounding)
What makes a true experiment?
A true experiment has internal validity
Reliability
The extent to which a measure is consistent (free from random error)
Ways to measure reliability
1) Test-retest reliability
2) Internal consistency
3) Inter-rater reliability
Test-retest reliability
If you measure the same individuals at two different points in time the results should be highly correlated
Internal consistency
Whether the individual items in a scale correlate well with each other – Cronbach’s Alpha assesses the correlation of each item with each other
Inter-rater reliability
The agreement of observations made by two or more judges
Ways to measure construct validity
1) Face validity
2) Content validity
3) Convergent validity
4) Discriminant validity
5) Predictive validity
6) Concurrent validity
Face validity
How obvious it is to the participant what the test is measuring
Content validity
Whether experts believe the measure relates to the concept being assessed
Convergent validity
The measure overlaps with a different measure that is intended to tap the same theoretical construct (the participant should be able to fill out two surveys and get correlating results)
Discriminant validity
The measure does not overlap with other measures that are intended to tap different or opposite theoretical constructs
Predictive validity
The measure’s ability to predict a future behavior or outcome
Concurrent validity
The extent to which the measure corresponds with another current behavior or outcome
Nominal scale
Numbers stand for categories but mean nothing themselves (male = 1, female = 2)
Ordinal scale
Numbers indicate rank order, indicating preference but not by how much (psych = 1, bio = 2, math = 3)
Interval scale
The distances between numbers on a scale are all equal in size, but zero is an arbitrary reference point (Likert scale)
Ratio scale
The only scale that measures a true amount of something. Zero means a non-existent amount of that variable, there cannot be negative numbers, and 4 is twice as much as 2
Close-ended question
Has a limited number of response alternatives, meaning higher specificity but less variety
Open-ended question
Allows respondents to generate their own answers, meaning more variety but less control and harder to analyze
Interview bias
The researcher may subtly suggest a desired response, interpret the response in the desired way, or probe open-ended questions to get the desired response
Respondent bias
Participants may act due to social desirability or response set (answering all questions similarly)
Ways to assess the construct validity of the independent variable
1) Pre-test
2) Manipulation check
Pre-test
Conducted before the actual study with a different set of participants and is meant to determine if the IV manipulation works as predicted
Manipulation check
Conducted during the study and assesses whether the manipulation of the IV had its intended effects
Ways to measure the dependent variable
1) Self-report
2) Behavioral
3) Physiological
Self-report
Asking participants about the behavior of interest; is easy and cheap, but is subject to bias
Behavioral report
Direct observations of participant behavior; is effective and direct, but can be expensive, time-consuming, and subject to reactivity
Physiological report
Directly recording responses of the body; is objective and measures strength of the reaction, but does not always capture valence and is subject to reactivity
Ways to control for participant expectations
1) Cover story: provides rationale
2) Filler items: reduces face validity
3) Placebo group: level of IV that shows role of expectations
Experimenter bias
When an experimenter might subtly suggest how they hope the participant will respond
Ways to reduce experimenter bias
1) Double-blind study: experimenter is blind to IV group of the participant
2) Blind to hypothesis: experimenter does not know the hypothesis of the study
3) Automated scripts and computers
4) Running participants in groups
Post-test only design
Participants are randomly assigned to one level of the IV and then measured
Pre-test Post-test design
Participants are given a pre-test and then randomly assigned to one level of the IV and measured
What is the purpose of a pre-test?
The pre-test gives a baseline measure of the DV before any IV manipulation in order to…
- ensure that groups are similar to start
- identify certain characteristics of participants
- measure the amount of change
- understand mortality
Between participants/independent groups
Each participant is randomly assigned to one level of the IV
Within participants/repeated measures
Each participant is assigned to all of the levels of the IV
What are the advantages of a repeated measures design?
- Participants are used more efficiently
- Can control for individual differences as each participant is their own control
What are the disadvantages of a repeated measures design?
- Could give away the nature of the study
- The order of presenting IV levels can impact results (control by counterbalancing and increasing time intervals)
Mixed factorial design
Combination of between participants and within participants
Matched pairs design
Order participants based on the independent variable, pair them in order, and randomly assign each pair to different groups
Factorial Design
Any experimental design with more than one IV
Need to consider main effects and interactions
Main effect
The direct effect of an IV on a DV. There is the potential for up to as many main effects as there are IVs
Interaction
When the effect of an IV on a DV depends on the level of another IV. There is a possible interaction for every combination of IVs
Moderator
An IV that affects the direction and/or strength of the relationship between another IV and the DV
Help us understand when an IV will impact a DV
Moderation exists when an interaction exists
“The effect of IV-1 on DV is moderated by IV-2”
Mediator
Represents the mechanism by which an IV influences the DV
Usually another DV that offers a deeper explanation for how the IV causes the main DV
IV –> mediator –> DV
Descriptive statistics
Statistics that describe the sample data
- measure of central tendency
- measures of variability
- distribution
- frequency
- correlation
- regression
- effect size
Distribution
General name for any organized set of data
Frequency
How often a score occurs
N
Sample size (# of data points)
Frequency distribution
Shows the number of times a score occurs in a set of data
Usually in a frequency distribution table
Bar graph
Used to demonstrate frequency of nominal or ordinal data
Histogram
Used to demonstrate frequency of interval or ratio data
Frequency polygon
Identical to histogram (frequency of interval or ratio data) but uses connected data points instead of bars
Mode
Most frequent score
Indicates central tendency with all scales including nominal scales
Median
Score than divides the group in have with 50% scoring above and 50% scoring below
Indicates central tendency with ordinal, interval, and ratio scales
Mean
Found by adding all the scores and dividing by the number of scores
Indicates central tendency with interval or ratio scales
Range
Largest value minus the smallest value in the sample – often inaccurate measure due to outliers
Standard deviation
Average deviation of the scores from the mean – more accurate since it uses every score
Variance
The standard deviation squared
Correlation plots
Measure the strength and direction of the relationship between two variables
Effect size
Refers to the strength of association between variables; provides a scale of values that is consistent across all types of studies (for example, Pearson’s r)
Pearson’s r
r = 0.15 –> small effect size
r = 0.3 –> medium effect size
r = 0.4 –> large effect size
r squared
Transforms Pearson’s r into a percentage of variance in one variable that can be accounted for by the other variable
Cohen’s d
Measures the standardized difference between two means
d = 0.5 –> the means are half of a standard deviation apart (medium effect size)
Simple regression
Predicts a score on one variable when the score on another variable is already known
Linear line of best fit through a scatterplot
Multiple regression
Used to combine a number of predictor variables to increase the accuracy of prediction of a given criterion or outcome variable (R)
Partial correlations
The correlation between two variables of interest with the influence of a third variable removed
Inferential statistics
Used to determine whether a sample of scores is likely to represent a certain population of scores
Based on the probability that the difference between means reflects random error versus real difference
Criterion
A value that tells us when we are going to decide a sample is too unlikely to have occurred through chance alone (alpha = 0.05)
Sampling error
A sample statistic that differs from the population parameter it represents due to chance factors
Type I error
The researcher rejects the null hypothesis when the null is actually true
Type II error
The researcher fails to reject the null hypothesis when the null is actually false (alternative is true)
Experimental realism
The extent to which experimental procedures have an impact on participants
Mundane realism
The extent to which experimental events in the controlled laboratory setting are similar to events which occur in the real world
Exact replication
An attempt to replicate precisely the procedures of a study to see whether the same results are obtained
Conceptual replication
Attempting to replicate the relationship between conceptual variables from the original study, but operationalizing these variables in a different way
Constructive replication
The replication wants to affirm the original research by fixing some methodological problems
Destructive replication
The replication wants to prove that the original research was wrong due to methodological problems
Advantages of meta-analysis
- precision
- objectivity
- replicability
- ability to make corrections
Disadvantages of meta-analysis
- statistics over reason
- objectivity and replicability can vary
- significant vs. practical
Experimental research
Explaining behavior by determining cause and effect relationships among variables
Correlational research
Looking for relationships among variables
Descriptive research
Making observations that describe behavior
Advantages of descriptive research
- higher external validity
- higher construct validity
- higher mundane realism
Disadvantages of descriptive research
- lower internal validity
- lower reliability
- lower experimental realism
- potential for observer bias
Observational research
Describing behavior
Naturalistic or systematic
Naturalistic observational research
The researcher makes observations in a natural, social setting
Qualitative: a small sample described in great depth (consider participant or nonparticipant and concealed purpose or not)
Inductive: begins with observations and generates hypotheses
Systematic observational research
The selection, recording, and encoding of natural behaviors
Quantitative: operationalize construct, determine setting and mode of observation, select sampling strategy, and train observers
Deductive: have a theory from which we generate hypotheses and use data to test hypotheses
Archival research
Using previously compiled information to answer research questions (statistical records, survey archives, written records)
Advantages of archival research
- free or cheap data
- abundance of data
- span of time periods
- look at reactions to natural events
Disadvantages of archival research
- low internal validity
- low reliability
- biases/errors
- no ability to gather extra information
Confidence interval
The range of scores around the sample results within which you have confidence that the true population value lies (allows generalization)
Probability sampling
Each member of the population has a specified probability of being included in the sample
- simple random sampling
- stratified random sampling
Simple random sampling
Every member has an equal chance of inclusion in the sample
Stratified random sampling
Subgroups are chosen and then random sampling occurs within those subgroups
Non-probability sampling
We don’t try to accurately represent the entire population within our sample
- convenience sampling
- quota sampling
Convenience sampling
Using the most convenient participants for your sample
Quota sampling
Choose subgroups and then use convenience sampling within the subgroups
Quasi-experimental designs
- one-group pretest-posttest design
- nonequivalent control group design without pretest
- nonequivalent control group design with pretest
One-group pretest-posttest design
Participants are tested on a quasi-experimental DV before and after the application of the IV (only one level of IV)
Nonequivalent control group design (without pretest)
Participants are assigned to an IV level based on an established variable, undergo application of IV, and then have the DV measured
Nonequivalent control group design (with pretest)
Participants are assigned to IV levels based on an established variable, pretested on the DV, undergo application of the IV, and then posttested on the DV
Interrupted time-series design
Multiple measurements of the DV occur before and after treatment
Interrupted time-series design with nonequivalent control group design
Two interrupted time-series designs are conducted with one group receiving a treatment and one group not receiving a treatment
Developmental research designs
- cross-sectional method
- longitudinal method
Cross-sectional method
Persons of different ages are studied at one point in time
Longitudinal method
The same people are studied at different points in time as they age