Quants Flashcards
First Quant textbook released in 1938
Experimental Psych - Robert Session
Purpose of experiment?
To demonstrate causality
3 requirements for causality?
- Temporal Precedence
- correlation
- No confounds
Early psych research was linked to what kind of science?
hardline sciences, more quantitative, focused on proving
Best friend in research
Control
Three key features of an experiment
E.C.M
Establish IVs - create conditions that are under control of researcher
Controlling Ext Vrs - avoid confounding
Measuring dep vars - must be precisely defined
Minimum no. of levels for IVs?
2
List 4 categories of IVs
- Manipulated or subject Vs
- Situational
- Task
- Instructional
What do we need to control?
EVERYTHING that is not of interest to us
How could one control for observer expectancy?
A double blind study
What is a vital element in establishing credibility in an experiment
The operational definitions
Should op defs be based on context, culture, times etc?
Yes
List 4 types of validity in experimental research
(ICES)
- Internal
- Construct
- External
- Statistical conclusion
Components of Stat conc val? (4 points)
- Correct analysis, no violation of assumptions
- report all analyses
- don’t fish for any and all sig analysis result
- use reliable measures (so you can find effects)
Two questions surrounded construct validity?
- Do your test measure what it claims to?
- Are your operational definitions adequate?
Central issue around external validity
GENERALIZABILITY
- across pops
- across environments
- across time
Definition of Internal validity
The degree to which a study is methodologically sounds and confound free
General process of exp studies?
OXO
Threats to Internal Validity
HMTSIAR
HOW MANY TEA SPOONS IN A REACTOR
(History, Maturation, Testing, Selection Bias, Instrumentation, Attrition, Regression to the mean)
History
what happens between the pre- and post-test that is outside of the study (war, protest, life change, retrenchment)
Maturation
Improvement or decline in abilities due to age/maturation (mostly bio/physiological, and mostly relevant to young/old pops)
Testing (threats to Int Val)
And possible solution
process of pre-test (rather than intervention) that affects change seen in post-test. (learning skills, gaining insight into own lives, asking big questions of their own life etc)
Distinction between testing and history/maturation is that is was the process of the pre-test which stimulated the change.
Possible sol: use a distractor
Instrumentation (threats to int val)
Lack of reliability and poor op defs introduce confounds (decrease int val)
Sol: Have good Rel and Op Defs
Selection bias (threats to IV)
And possible solution?
Sub-con selection of people who may respond better.
Sol: RA (needs LS)
Reg to mean (threats to Int Val)
And possible solution:
V high/low scores in pre-test and normal scores in post-test (eg nervous in pre-test and calm in post). Seems like intervention is effective or not when it’s the opposite.
Sol: LS
How could one protect against history?
Pre- and post-test in a single sitting. Or, try collect data on what happened to participants between tests.
Name two basic experimental research designs and briefly explain them and one problem with each
- Between subjects - each P takes part in only one condition of research. Problem: Creating equivalent groups
- Within-subjects (repeated measures) - each P in all conditions of research. Problem: sequencing effect (having been in one condition may affect behav in other condition
When should one use between subject designs (2 marks)
- When IV is a Subject variable (eg intro/extrovert). things researcher cannot control
- When sequencing effect prevent within-subject design (eg given coping skills one can’t unlearn)
What is needed for groups in between-subject designs, and list two ways to make the groups so
Equivalence.
1. Random assignment
2. Matching
Benefits of random assignment in Between-subjects design experiments (4 points)
Helps nullify:
- Bias selection
- History
- Maturation (control group controls for this)
- Implementation
Discuss matching in creating equivalent groups in between-subjects design (4 points)
- try make sure everyone is equal on things that are linked to DV (eg get people with brain lesions and group them according to what lesions affect)
- Not as good as RA
- not a true exp with matching
- useful when only a few Ps are available (no RA possible)
Advantages of within-subjects/repeated measures designs (3 points)
- Need fewer people
- no worries about equivalent groups
- decrease error variance (no indi diffs affecting results)
Problems with within-subjects/repeated measures designs (3 marks)
- Carry-over effects (does it matter if condition A or B comes first?)
- fatigue effects
- practice effects
Explain factor vs level
Factor = the IV
Level = number of “states” of that variable you’re testing
Indep groups, 1-factor
groups of cats in 2 visual environments, horizontal vs vertical stripes. Factor and Level?
Factor = vis enviro
Level = vertical or horizontal stripes
Matched groups, 1 factor
Effect of sleep dep on influence of misleading questions? Factors and Levels
Factor = sleep deprivation
Level = no. of hours of sleep dep
Nonequivalent groups, 1 factor
Are gifted children good at emotional problem solving compared with average IQ kids.
(non-eq groups as Ps can’t be randomly assigned)
Factors and Levels?
Factor = IQ
Levels = gifted or average IQ
Within subjects, 1 factor
Will kids shift balance to moving stimuli as if their balance has shifted?
Factors and Levels?
Factor = visual stimuli
Levels = Forwards and backwards
Between-subjects, multilevel designs
Effect of caff on reaction time (varying levels of caffeine exposure)
* NB to looks at enough levels, as relationship may not be linear
Factor and level?
Factor = Caffeine
Level = amount of caffeine
2 factor study with 3 levels is called a…. and has … conditions
3x2 factorial design and has 6 conditions
main vs int effect and eg
(multifactorial designs)
Main: effect of one IV on DV
eg: caffeine affects word finding ability
Interaction effect: does effect of one IV depend on level of another IV?
eg: effect of caffeine on word finding ability is dependent on level of sleep deprivation
In multifactorial design:
For main effects, look for significant differences in the….
For interaction effects, look for significant differences between the…
…overall averages
…cells of a given row/column
In multifactorial design, a condition is a…
…subgroup created by different factors (eg males with high anx)
Can factors in a multifactorial design study be both 1 between-subjects and 1 within-subjects?
Yes.
Eg, spider approaching or retreating (one factor, within-subjects) AND self-efficacy hi or low (between subjects)
as you number of conditions go up so do your….
number of required participants
The two broad disciplines of scientific psych?
- Correlational psych - concerned with indi differences and relationship between naturally occuring variables
- experimental psych - no interested in indi diffs, but minimizing them for generalizability’s sake
Correlation vs regression?
Correlation = Identifies association between two variables
Regression = used to make predictions when strong correlations exist (Y=a+bx)
Problems with correlational research and causality (2 marks)
- Directionality (temporal prec)
- Third variables (confounds)
Solution to directionality problem in correlational research
Cross-lagging - for studies over time (reports on strength and relation)
Can one infer the design from statistics?
No
Why do we do correlational research despite short comings? (3 marks)
- Practicality (some variables can’t be randomly assigned - gender, age etc)(but can find ideal participants)
- Some R is conducted with prediction in mind
- Ethical grounds (can’t randomly assign brain damage)
4 places where correlational research is used
- Psych testing
- Research in Personality and Abnormal Psych
- Studying nature vs nurture controversy
- Any cross-sectional study
Most correlational research takes place where?
Outside the lab - quasi exps and programmer evaluation.
Correlational R used when …. are not possible
experimental procedures
A true experiment has….(2 marks)
- random assignment
- more than one condition
Quasi-experiments are those in which participants…..
cannot be randomly assigned
Main advantage of quasi-experiments (1 mark)
Increased external advantage (done in real word)
Two types of quasi-experimental designs?
- Nonequivalent control group designs
- Interrupted time series designs
When does one use a nonequivalent control group design?
When one cannot properly randomly assign
Interrupted time series allows for…and is used when
the evaluation of trends….there is no other way to find the answers
Should one take multiple observations before and after the interruption in an interrupted time series design
Yes - observations on either side of the interruption may be part of a bigger trend
Cannot have RA in a quasi experiment but one can have a….
..nonequivalent control group (town in drunk driving example)
Where are small N designs most useful?
Clinical populations
Why use a small N design?
- subjects are hard to find (rare diseases etc)
Requirements in a small N design (3 marks)
- target behavior operationally defined and easily measurable
-establish a baseline of responding to intervention
-introduce treatment and keep on monitoring
Small N studies must be ….. as they can form the base/pilot for future research
reproduceable/replicable
Depending on how much control the R’r has over IV, small N studies can be…
quasi-exps
Two types of small N designs:
- Withdrawal (A-B; A-B-A; A-B-A-B max)
- Multiple baseline design
When are multiple baseline designs used?
When withdrawal studies aren’t possible (intervention not removable, or unethical to remove)
In a multiple baseline design, introduce intervention to participants at same time?
No, at different stages
Evaluating small N designs:
External validity?
Stat significance?
Interactions?
Studies that don’t use frequency?
Ext val generally low
No stat sig due to low numbers
Not much info in interactions from small N’s
Need studies that use frequencies
Name 2 NB aspects of surveys
- Sampling
- Structuring the questions
Probability sampling aspects: (3 points)
- Each person has an equal chance o being chosen for sample
- Learn something specific about an identified group of indis (generalise back to pop)
- Representative sample needed
Three types of random sampling and explanation
- Simple random sampling - good, except when wanting to study systematic features of pop or pop is very large (might not get fair rep of all departments in uct etc)
- Cluster sampling - Do simple random sampling within each cluster to ensure equal rep
- Stratified sampling - looking at a particular characteristic, but cluster does not exist. Simple R Sampling from each strata
Non-prob samplin: When to use it? (1 point)
When generalizability is not the goal, but the relationship between the two variables is
Non-prob sampling: two types
- Convenience sampling - take whoever responds to request and who meets general requirements
- Purposive - very specific seeking of Ps
Survey methods - 3 types
- Interview
- Written surveys
- Phone surveys
Advantages of interviews (surveys)
- comprehensive, lots of data
- few problems with unclear info/missing data
Disadvantages of interviews (surveys) (4 points)
- People refusing to be interviewed
- cost, time, logistics (travel, transcription etc)
- interviewer bias (cross-race bias etc)
- Hard to standardize if using R assistants.
Written surveys use either… or …. questions, and sometimes….
open, closer, likert scales
Administration was challenging before internet
Phone surveys disadvantages: (2 points)
- Excludes pop without phones
- non-response rate can be high
Things to be cautious of in surveys: (4 points
- Sampling frame bias
- Social desirability bias
- Survey too long
- Bad questions (leading/2-part)
4 Types of observational research
- Case studies
- Naturalistic observation
- Participant observation
- Archival research
Two approaches held in tension in Case Study research, and their focus
- Ideographic (focus on individual)
- Nomothetic (focus on group)
Which type of observational research allows for study of rare phenomena
Case studies
Which type of observational research provides a source for developing ideas, hypotheses and therapy techniques
Case studies
Which type of observational research provides very persuasive data but have low generalizability
case studies
limitations of case studies (2 points)
- rely on anecdotal info, this makes them vulnerable to bias presentation
- low generalizability
Which mode of observational research attempts to study normal behaviors of people or animals in their environment
Naturalistic observation
(and participants observation, but here the researcher joins the group being observed)
Shortcomings of Observational research (3 marks)
- no control group; no causality can be drawn
- observer bias; preconceived ideas shape interpretations of people’s behavior
- subject reactivity (behave differently when their observed)
Type of observational research that draws from data ranging from public records (census data, court records) to personal info (credit rating, diaries etc)
Archival research
Advantages of Archival research (3 points)
- unlimited info available
- can mix with lab research (increasing external val)
- no subject reactivity
Disadvantages of archival research (2 points)
- info may be missing or not representative
- experimenter bias (selecting only records that support one’s hypothesis)
Which study and report began the formalization of ethics
- The Tuskegee syphilis study
- The Belmont report
Name three fundamental principles to ethics
- Beneficence (maximise benefits and minimise risks)
- Autonomy (people decide whether or not to participate in research)(Deception makes this tricky - just make sure there’s no hard to participants and they are debriefed after participation)
- Justice (fairness - who to include in the study, who bears the risk in the research, who reaps the reward from the research)
Boards and codes of ethics (3)
- Institutional review board (IRB)
- The APA ethics code
- The Helsinki declaration of the World Medical Association
Three components of informed consent
- Knowledge (understand nature of exp, alternative available, risks and benefits)
- Volition (provide consent free from constraint of duress - may revoke at any time)(Don’t offer large cash incentive in low SES comm - no one will be able to say no)
- Competence (indi is able to make well-reasoned decisions and give meaningful consent)
*don’t confuse these with fundamental principles to ethics
3 other ethical issues in R
- Fraud (do not make up data)
- Allocation of credit (do not plagiarize)
- Sharing of materials and data (share materials and data so people can run different analyses)
Components of consent form
- Overview
- Procedures
- Risks
- Benefits
- Cost/economic considerations
- Confidentiality
- Alternative treatments
- Voluntary participation
- Questions and further info
- Signature lines
What is counter-balancing?
Attempt to mitigate sequencing effects in a repeated measures study - one half does A then B, other half does B then A