1/112
Looks like no tags are added yet.
DV
What you measure
Operational definitions
Definition of a theoretical construct that is stated in terms of concrete observable procedures
Example —> aggression as how much hot sauce participant asks another to eat
Reliability
How consistently your measure assigns the same number to the same observation
If a ruler says an 8-inch paper is 9.32 inches every time, it is reliable
Validity
How well your measure actually measures the thing it's supposed to measure
Ruler tells you 8 inch paper is 8 inches long, it is valid
Can’t be valid unless reliable
Convergent validity
Measure your construct in several different ways in addition to the way you want (see if it correlates with other things it should correlate with)
Measure of optimism should correlate with measures of optimism
Discriminant validity
Make sure construct correlates much less with constructs not related to it
Measure of optimism should have smaller correlations with measures of happiness, positive mood, and hopefulness
Types of measures
Self-report, behavioral , implicit, and physiological measures
Self report measures
Can use to measure sexual behavior and condom use (Viagra was approved just based on self-report), fill out questionnaire
Self report measures advantages
Easy to use, certain directness, lot of qs in a short time, ask some qs that are there solely to help you untangle confounds later, can get subtle changes in participant’s feelings, questionnaires can be anonymous so people are more likely to admit private or embarrassing things
Basically lots of info quick and easy
Self report measures disadvantages
Subjects may treat them casually, may try to make themselves look good, not be honest, act based on expectancies or reactance, can tell you are assessing them, responses can be influenced by trivial things (anchoring effects, question order [asking how satisfied you are about your life and then job, if switch the order might answer life lower because thinking about bad job], context [ask how trustworthy leaders are after ask how trustworthy putin is])
Info can sometimes be untrustworthy
Behavioral measures
Give subjects an opportunity to perform a behavior and see if they do it (conformity), how much (how many shocks given/following orders), long (task engagement/time kids play w markers), and many times (attraction, times smile at each other)
Unobtrusive measure
Type of behavior measure that is made in absence of subject —> how much dust left on books in library to see use, how big tombstones are to show wealth, content analysis of news coverage and online threads
Behavioral measure advantages
Measure a behavior without a subject knowing, more engrossing than survey so more true response, behavior may be closer to construct of interest than questionnaire
Basically can trust responses
Behavioral measure disadvantages
Harder to do than questionnaires, sometimes isn’t possible (measure condom use), people may do same behavior for different reasons (agree to date based on liking or pity), takes a lot of time to get behavioral measure (one subject at a time), can’t measure many things at once
Harder to do this measure, sometimes meaning of behaviors unclear
Implicit measures
Measures of people’s unconscious, automatic attitudes gotten without asking directly
Often used when assessing sensitive social issues (prejudice; immoral behavior)
Examples → priming
Implicit measures advantages
Not influenced by self-presentation concerns, don’t require people to have accurate self-knowledge, doesn’t matter if subjects are unwilling to be honest with themselves
Subjects can’t control responses
Implicit measures disadvantages
Not always clear what is actually measured (fail to correlate with self-report or behavioral, multiple implicit measures of same construct don’t correlate much), sensitive to context
Not always clear what responses mean
Physiological measures
Sample some kind of physiological output (heart rate, cortisol, inflammatory activity, blood flow to different parts of brain, gut microbiome)
Physiological measures advantages
Subjects don’t have control over variables, easy to measure
Can’t control responses
Physiological measures disadvantages
Can scare people (blood samples), can be expensive, multiple physiological measures of same construct don’t always correlate
Techniques scary and expensive
Solution for disadvantages of DV measures
Have a little of every single type
Quasi-experiments
Can’t have random assignment in all conditions, so can’t say anything about causality
Experiment comparing people with pro-israeli or pro-palistinian views and see same news clip to see attitudes (can’t assign if they are pro-israelie or pro-palestinian)
One way two groups designs
One IV, two levels of IV (job applicant or patient)
One way multiple groups designs
One IV, more than two levels of IV (overjustification study → played with markers to win an award, just played, played and unexpectedly won an award)
Factorial designs
Two or more IVs —> Variables are completely crossed → every level of every IV appears in combination with every level of every other IV
3 x 4 design, one IV has three levels other has four
If theory suggests that two variables tend to occur together, but you want to see separate combination of each variables (or combined effect)
2×2 designs
Two IVs, each has two levels
Dieters vs non dieters → either given two milkshakes or none (ask about flavor but actually measure how much is eaten)
DV is how much they ate, operationalized as difference in weight after compared to before (eat ice cream after milkshake)
Person by situation quasi-experiment
Main effects
Effects of one IV
Did dieters eat more ice cream than non dieters?
No effect because average of all conditions is equal
Did people who had no milkshakes eat more ice cream than people who had 2?
No effect because they are equal
Interactions
Effect of 2 IVs together
Dieting status and preload size (how many milkshakes) interact to affect how much ice cream is eaten
If you’re a non dieter, drinking two milkshakes causes you to eat less ice cream, but dieters are opposite
Do lines cross on a line graph? Then interaction between the two
Simple effects
Comparing one condition to another
Between subjects
Each subject is exposed to one and only one of the experimental conditions
When we compare the conditions we are looking at entirely different people
Within subjects
Each subject is exposed to all experimental conditions
When we compare conditions we are looking at all the same people
Between design adds random error → if someone in the study is stubborn in an attitude experiment, show how they have low scores in both conditions (controls for individual differences)
Fewer subjects needed
Contamination effects (within subjects)
Improvement/or worsening with practice/time
Interference effects
When the approach that worked one day doesn’t work the next day
Order effects
Something about the passage of time affects how participants respond (boredom, hunger, fatigue, etc)
Counterbalance the order of conditions → vary the order in which subjects participate in the different conditions of the study (only works for reversible manipulations, not like positive vs negative feedback from personality test)
Complete counterbalancing
Present every possible order of experimental conditions to the subjects
2 conditions, A and B, there are 2! Orders (2 x 1): AB and BA
4 conditions (4 x 3 x 2 x 1): 24 orders, so gets worse with more conditions
Latin square
Solution for counterbalancing problem, each condition appears exactly once in each position so there are as many orders as conditions
Order 1 → a is first, b is second, c is third, d is fourth, a is never first again
Mixed model
Have more than one IV (one is between, other is within; between-within design)
Safe vs scary imagery of spiders, with spider moving or still
Make 1st between, 2nd within
Mixed model problems
Contamination effects → subjects get scared in one condition, they could stay scared
To correct this → could have a cool-down session between videos, counterbalance the order of the videos (of spider moving)
Standard deviations
Average distance between scores in sample and the mean
95% of items within 2 SDs of mean
When trying to figure out if observation is unusual, need to look at SD and mean
Normal distributions of human psych and behavior don’t exist irl
ANOVA equation
F = (difference groups)/(normal variability within groups)
If f greater than four, differences between groups much bigger than normal variability within groups (two groups are importantly different from one another)
If f is less than one, have opposite effect
ANOVA def
T-test comparing means of multiple groups on continuous outcome variable
2 x 2 → smiling and providing good service producing better tips
Look at all main effects and interactions between variables (example of when used)
Pearson correlations
Test magnitude and direction of association between two continuous variables (variable that has 3+ scale points, not either a 1 or 2 as an answer)
Income and years of experience (r = 0.86, strong and positive)
Example of when used → association between loneliness and gpa, living in dorms or off campus feel more lonely?
T-tests
Compare two different groups
Independent samples t-test
Between subjects,
Does group a score better than group b?
Paired samples t-test
Within subjects, does group a improve from test 1 to test 2?
Example of when used → did an intervention work to improve feelings of loneliness from time 1 to time 2?
Chi-square test
Compares expected and observed frequencies of categorical outcome variables (conform/don’t conform, cheat/don’t cheat)
Either democrat or republican, either vote or not vote on an issue
Expect 25% for each condition, but doesn’t turn out that way
Example of when used → sd natives less likely than non-natives to seek counseling? (yes or no answer for dv)
Suprise index (chi-square)
χ2 stat indicated how far observed
Animism
The belief that natural phenomena are alive and influence behavior (sun has wills and temperaments)
Metaphysical system
Posivitism
Theories of knowledge and human behavior should be based only on observations that can be made with absolute certainty (so not studies of god; if you’re
going to study something pick a topic where you can make real observations)
Determinism
The notion that the universe is orderly and that all events have meaningful, systematic causes (people tend to think causally—illusory correlations,
superstitions)
One of four canons of science
What makes a study important?
Challenges existing theories, integrates existing theories (do people do worse or better with an audience —> better when they know what they’re doing, worse when they don’t), challenges institutions or expectations, results in a large effect from a small manipulation, speaks to issues of societal importance (SES health gradient)
Motivations for doing research
Personal interest, observation (ww2 atrocities), previous literature, practical problems (how to help others)
4 goals of psych research
Describe behavior (what happens), predict behavior (when does it happen), explain behavior (why does it happen), change behavior (what changes what)
Philosophy def
Study of knowledge, behavior, and the nature of reality by using logic, intuition, and observations (competes with metaphysical systems)
Empiricism
An approach to understanding the world that involves collecting data or making observations (“got it straight from the horse’s mouth”)
One of the four canons of science
Parsimony
Should prefer simple explanations over complex ones
One of four canons of science
Falsifiability
Openness to disconfirmation; scientists should seek out tests that could prove their theories wrong (Popper –logical positivism)
Part of testability which is one of four canons of science
Operational defs
Definitions of theoretical constructs that are stated in terms of concrete, observable procedures
Four ways of knowing about the world
Intuition, logic, authority/expertise, observation
Science (all last four)
Law
Universal statement of the nature of things that allows reliable predictions of future events
Theories
General statement about the relation between two or more variables—should be deterministic, logical, orderly, emphasize causal relations between variables, empirically grounded (testable), parsimonious
Different than laws because only happen in specific circumstances
Hypotheses
Predictions about specific events that are derived from one or more theories
Method of induction
Reasoning from specific instances to general principles (see lots of individual birds fly, come to the conclusion that all birds can fly)
But when do you know if that is actually the case (fundamental attribution error)
Method of deduction
Reasoning from the general to the specific. Occurs when a theory is used to develop predictions that are then tested against observations.
Positive test theory
Tendency for people who are evaluating hypotheses to attempt to confirm rather than to disconfirm these hypotheses
Behavioral confirmation
Tendency for social perceivers to elicit behaviors from a person that are consistent with their initial expectancies of the person
Three approaches to hypothesis testing
Validation (support or confirm a theory), falsification (disconfirm a theory), qualification (try to identify the boundary conditions under which a theory or
hypothesis is and is not true)
Methods of developing ideas (inductive)
Case studies, paradoxical incidents, analyzing things that experts in a particular field do to achieve certain outcomes, luck of fortune (notice something important by accident)
Methods of developing ideas (deductive)
Reasoning by analogy (analogy between biological resistance to disease and psychological resistance to persuasion; thermostat as a metaphor for understanding self-regulation; using pain system to understand rejection), applying a functional or adaptive analysis (what do organisms have to do to survive), accounting for conflicting results/exceptions, hypothetico-deductive method (begin with a set of basic assumptions and derive one or more logical consequences from these basic principles —> people like predictability, so want people to see them as they see themselves)
3 requirements to show one thing causes another
Covariation (change in one thing changes another), temporal sequence (change in the first variable must precede change in the other), eliminating confounds (being able to conduct a true experiment [eliminating third variables])
Non-response bias
Because only 25% of people returned surveys, non-respondents may have different preferences from respondents
Selection bias
Survey systematically excluded voters who were poor (likely to be democrats) by using subscriber lists, phone books, and car registries
Construct validity
Extent to which the IV and DV in a study truly represent the abstract, hypothetical variables of interest to the researcher
Conceptual validity
How well a specific research hypothesis maps onto the broader theory that it was designed to test
Reliability/test-retest reliablity
Consistency or repeatability of a measure or observation, see if someone gets the same score on the same measure with some time apart
Internal consistency
The degree to which the total set of items or observations in a multiple-item measure behave in the same way
Interobserver agreement/interrater reliability
Degree to which different judges independently agree upon a measurement, judgment, or observation
Correlation fallacy
Assuming causality from correlation, CORRELATION ≠ CAUSATION
To show causality and generalize
Need to hold all IV’s constant except those of interest, randomly assign subjects to each experimental condition, generalize by randomly sampling from population of interest
ONLY CAUSALITY IF RANDOM ASSIGNMENT
IV vs DV
IV —> variable you manipulate
DV —> variable you measure
Internal validity
Did IV actually influence DV
External validity
Can you generalize your findings
Random error
Error caused by irrelevant variables whose average influence on the outcome of an experiment is the same in all conditions (variables that you didn’t control for that affect everyone in the same way)
Doesn’t affect internal validity, but can hide effect of IV —> keep as small as possible
Systematic error
Error caused by extraneous variables that tend to influence all scores in one condition and to have no effect/different effect on scores in other conditions (threatens internal validity —> goal to eliminate it)
History threat
Specific events that happen during the course of a study can affect the variable being measured (kid’s fear of swimming scores increased not because of watching scary movies but because of a tsunami near them)
Need to have a control group
Maturation threat
Changes that happen within a person due to time passing can affect the variable being measured (participants in a coping study may learn to cope better over time even without an intervention)
Need to have a control group
Regression to the mean threat
The tendency for people who receive high or low scores on a particular measure to score closer to the mean on subsequent tests (if people take the SAT and do horrible, will do better next time regardless of if they get help, will score closer to mean score)
Need to have a control group
Don’t select people based on extreme scores
Testing effects threat
People tend to do better on a test the second time they take it (people do better on the SAT the second time regardless of if they take a prep course or not)
Have a control group
Don’t give a pretest
Use a different type of test the 2nd time
Experiment mortality (heterogenous attrition) threat
Different amounts/type of people drop out of each condition of your experiment (heavy smokers drop out of intervention study, so looks like intervention worked even if it didn’t)
Minimize drop outs (increase retention)
Compare the drop-outs from the conditions
Participant reaction biases threat
Participants try to behave in ways that are consistent (or not) with researchers’ hypothesis
Subject expectancies and reactance
Give a false hypothesis
Convince participants they are an experimenter
Keep participant unaware they’re in a study
Use behavioral measures instead of questionnaires
Have a non-obvious hypothesis
Subject expectancies
Subject acts in ways that are consistent with researcher’s hypothesis [demand characteristics] (participants are told that they’ve been given alcohol, act drunk even if they’re not)
Subject reactance
Subject acts in ways that aren’t consistent with researcher’s hypothesis
Experimenter bias/influencer threat
Experimenters’ expectancies can affect either what they observe or how they act (if experimenter thinks the participant has been given alcohol, more likely to notice if they slur their speech)
Keep E blind to hypothesis
Keep E blind to what condition P is in
Have two E’s that are each partly blind
Have all instructions pre-recorded
Confounds
Some additional variable (that you don’t care about) varies systematically along with the thing you manipulated (alternative explanation) —> compliment from parents vs no compliment (interaction w parent could be positive)
Do a different study to rule out confounds
Measure variables that might be confounds
Add a condition that helps you isolate the additional variable
Homogeneous attrition threat
Particular type of person drops out of your study, regardless of condition (all of heavier folks drop out of dieting study regardless of condition)
Minimize drop outs
Increase external validity
Real-world realism (mundane realism) —> studies that look like the real world (studying sleep behavior by turning lab into a bedroom)
“Psychological” realism (experimental realism) —> psychologically meaningful to the subject or feel like the real-world (asch’s conformity studies)
Threats to internal validity
History, maturation, regression to the mean, testing effects, experiment mortality, participant reaction biases, experimenter bias/influence, confounds
Threats to external validity
Homogeneous attrition
Environmental manipulation
Manipulation of physical setting
For studying memory —> change scent to be unfamiliar or familiar