psyc 270 learning check 2

Correlational Design

-An indicator of the strength of association between two variables -Range: -1 to 1 -Positive: height and weight -Negative: # alcoholic drinks and reaction time -Zero: gender and IQ

Internal analysis

-Break treatment group into one or more subgroups -Test for subgroup differences that are consistent with the favored theory or with competing theories

Time-Series Design

-Collect multiple waves of data from two presumably comparable groups -If similar scores on DV BEFORE manipulation but different scores immediately AFTER, strong evidence manipulation may be responsible

Two Group, Pretest-Posttest Design

-Compare two groups both before and after one group receives a natural manipulation -Strong quasi-experimental design IF we can assume groups were similar before manipulation

Factorial Designs

-Contain two or more IVs that are completely crossed -Crossed: Every possible combination of all the levels of all IVs -2 X 2s are quick and informative -Allow qualification: theory is true under some conditions but not others -Qualification is an interaction

Archival Research

-Examine naturally existing public records to test a theory or hypothesis -Hospital records, marriage licenses, political speeches -Sidanius & Pratto (2000) analysis of death penalty cases

Design

-plan for the structure of a study -Tells us what will be done to whom and when

5 types of non-experimental design

1) Archival research 2) Case studies 3) Single variable studies 4) Observational research 5) Correlational studies

Conceptual Validity

-How well a specific hypothesis maps onto the broader theory -Hypothesis should --Follow logically from theory --Be tough to generate without theory

Survey

-Identify a subset of people in the population to study -Use their answers to estimate answers of the entire population

Ratio

-Like interval scales except has a true zero point -A point at which none of the quantity under consideration is present -Can never have a negative value

One way, multiple groups design

-One IV -IV has three or more levels --Prime: collectivism, American, Individualism, Self, No prime/ neutral prime

Two groups design

-One IV -The IV only has two levels --Experimental (e.g., collectivism prime) --Control (e.g., no prime)

External Validity

-a.k.a. generalizability -Extent to which research findings provide an accurate description of what happens in the real world --To people --To situations

two challenges of survey writing

-judgement phase -response translation phase

Operational confounds

A manipulation influences more than one psychological construct, each of which may influence the DV

Manipulation check

A measure designed to see if a manipulation truly puts people in the psychological state that the experimenter wishes to create

Patching

A researcher adds new conditions to a study to -help establish the size of an effect -test for the influence of confounds -or both

Posttest-Only Design

Compare two similar but nonidentical groups after one and only one of the groups experiences a treatment -2 sections of Intro to Psych --One section completes collectivism word-find puzzle, the other does nothing --Measure prosocial behavior

Third variable

Confound that usually exists in passive observational studies

Nominal

Categorical Meaningful but potentially arbitrary and non-numerical names

Covariation

Changes in one variable must correspond to changes in another variable

two functions of a good design

Enables us to pinpoint effects of the IV on the DV Helps us rule out confounds

noise

Extraneous variables that influence the DV but are evenly distributed across experimental conditions

Internal consistency

Degree to which the total set of items or observations in a multiple-item measure behave in the same way

Single Variable Research

Designed to describe some specific property of a large group of people Typically descriptive research Usually cannot test causality

Factor Analysis

Determines if all items on a measure assess the same psychological construct or if some items are influenced by a different construct(s)

non-experimental design

Does: -Attempt to understand an interpret behavior Does not typically: -Manipulate a variable -Randomly assign participants to condition -Have enough control over situation to decide who receives a treatment or at what time -Search for the cause of a behavior

Mundane realism

Make a study as similar as possible to a real world setting that you care about

Case Study

Making careful analyses of the experiences of a particular person or group Try to explain unusual events by relying on established scientific principles

One-Group, Pretest-Posttest

Measures are taken from a single group of participants both before and after a natural manipulation

Median split

Problem: People near cutoff (barely below or above median) are much more similar to one another than they are to the extreme members of their own group

Bipolar scale

Rate a quantity that deviates in both directions from a zero point

Unipolar scale

Rating begins at a very low value (e.g., 0) and moves up to a subjective maximum point on the dimension of interest

Sampling error (a.k.a. margin of error)

Reflects likely discrepancy between the results we get in a specific sample and results we probably would have gotten from the entire population

Interval

Interval Make use of real numbers designating amounts to reflect differences in magnitude Sometimes has a negative value Unit corresponds to a specific amount of the construct being measured

Observational Research

Investigators record the behavior of people in their natural environments Q: Do students demonstrate more group pride after their school succeeds at athletics compared to after it fails? A: Observe students leaving dorms. Record number wearing UNC gear after a) Lose - b) Win +

Ordinal

Involves order or ranking Provides information about relative Not sensitive to absolute differences between these things

Independent Variable

Variable that is manipulated by the experimenter

Dependent Variable

Variable that is measured by an experimenter

Procedural confounds (a.k.a. environmental confounds)

Something the experimenter did not intend is allowed to covary with the IV

Test-retest reliability

Testing a group of individuals at one time and then having them come back a second time to take the test again

Validity

The relative accuracy or correctness of a statement

random assignment

equate experimental conditions before manipulations are applied in experiment; internal validity

high internal validity

confidently conclude that variations in the IV caused any observed changes in the DV

3 steps of internal validity

covariation, temporal sequence, confound

random selection

increase likelihood that small sample generalizes to population in population surveys; external validity

Temporal sequence

to argue that changes in one variable Cause changes in another, changes in the first variable must precede the changes in the second

Response Translation Phase

Can participants translate internal psychological state into a value on response scale?

Artifact

-Variable held constant in studies that may influence research findings -"the lab" is an artifact

Patched up Designs

Quasi-experimental design plus control groups

Quasi-experiment

Research designs in which researchers have only partial control over their IVs -Participants are assigned to one or more conditions by some means other than random assignment -Typically fewer concerns about external validity

Interobserver agreement/ Interrater reliability

The degree to which different judges independently agree upon an observation or judgment

Internal Validity

The extent to which a set of research findings provides compelling information about causality

One-Group Design

-All participants are in a single group that received a natural or experimental manipulation -Q: Does adopting a collectivist mindset increase prosocial behavior? -All participants solve word-find with collectivist words

Experimental Realism

-Degree to which research is psychologically meaningful to participants -High experimental realism usually requires deception

Statistical Reliability

-Determines inter-item reliability within the scale -Some items are more reliable with their peers than others --higher item-level alpha = good

Nature and Treatment Designs

-Natural groups with experimental treatment design --Two naturally occurring groups are given different treatments --Useful if we can assume: ---Two groups were very similar in important ways before treatment ---Were exposed to different levels of experimental treatment while holding everything else constant Q: Does tax on bottled water reduce consumption across levels of income? Manipulated IV: bottled water tax (yes, no) Measured IV: income DV: consumption of bottle water

Open-ended questions

-Refine structured ratings scales -Use as primary source of data -Code in a number of ways -Coding is resource intensive -Reverts data to numbers that participants could have reported, if asked -May be vague, not provide much of an answer

Person-by-Treatment Design

-Researcher measures at least one IV and manipulates at least one other IV -Person-by-treatment quasi-experiment --Prescreening --Median split ---Problem: People near cutoff (barely below or above median) are much more similar to one another than they are to the extreme members of their own group

Reliability

-The consistency and repeatability of a measure or observation -More is better (less error) for observers, observations, and occasions -Reliability is a necessary but not sufficient condition for validity

Construct Validity

-The extent to which the IVs and DVs in a study represent the abstract, hypothetical variables of interest -Good operational definitions, high construct validity

One Thousand Lives a Month

-Trasylol -blood thinner linked to death -Case study: Randoni's surgery -Archival: hospital records -Single variable research: assess death rate after cardiac surgery when Trasylol was used as a blood thinner

Natural Experiments

-Uses naturally occurring manipulation -All variables are measured; NO random assignment -Focus on events that are fairly arbitrary, but do not occur completely at random -Q: Does water pollution influence support for penalizing companies that pollute? -Q: Does Plumpynut decrease malnutrition and death by starvation in young children?

Census

A body of data collected from every member of a population of interest

Confound

Additional variable that may (a) influence the DV and that (b) varies systematically along with the IV

judgement phase

Are participants thinking about the same question as the researcher?

Experimental Design

Goal #1: See if changes in IV lead to changes in DV Goal #2: Equalize experimental and control groups

Focus group

a small, representative, sample of participants from the group of interest meet to discuss their experiences