Research Methods Ch. 1-5

SciMeth Step 1

identify a problem

ScieMeth Step 2

develop a research design

SciMeth Step 3

conduct the study

SciMeth Step 4

analyze and evaluate the data

SciMeth 5

communicate the results

SciMeth Step 6

generate new ideas

Population

total people you want to study

Sample

people you end up observing

Goal of Sci: Describe

What is the behavior or event?

Goal of Sci: Explain

What are the causes?

Goal of Sci: Predict

Can we anticipate when the behavior will occur?

Goal of Sci: Control

Can we manipulate conditions to cause behavior?

Basic Research

asking questions to help us understand the nature of behavior

Applied Research

answers questions concerning practical problems

Quantitative Research

recording observations as numeric data

Qualitative Research

recording observations as non-numeric data

Theory

a broad statement used to account for an existing body of knowledge, as well as provide unique predictions to extend that body of knowledge

Hypothesis

testable claim or prediction about what you expect to observe

A good hypothesis or claim must be:

testable, replicable, or parsimonious

Deductive Reasoning

claim is used to generate ideas (test a theory)

Inductive Reasoning

observations generate ideas for the future (test an observation)

Three C’s of good lit-review (comprehensive)

read all titles, abstracts, intros, and discussion

Three C’s of good lit-review (critical)

ask yourself about limitations, methods, and results

Three C’s of good lit-review (clever)

look at the cracks between the data to come up with alternate explanantions or new research ideas

Publication Bias

the tendency for editors of journals to preferentially accept articles that show positive results and reject articles with negative results

Research Ethics

actions researchers must take to conduct responsible and moral research

nuremburg code

1st international code of ethics, as a result of nazi experimentation

belmont report

congress-backed, concrete ethical set of rules

Tuskeegee Syphillis Study

observation of poor black men w/syphilis to study long term effects

BR: Respect for Persons

autonomy and informed decisions

BR: Beneficence

minimize risks and maximize benefits

BR: justice

fairness and equity in participants of the study

Robbers Cave Experiment

two groups of boys in competition with each other in order to understand and manipulate prejudice

Ethics of Robbers Cave

fostering prejudice, conflict and aggression in kids, fixed by eliminating prejudice at the end with a collaborative activity

Milgrams Obedience Experiments

participants were expected to shock a confederate at increasing levels until a lethal point, to see how far obedience to an authority figure goes

Ethics of Milgrams Obedience

stress on participant (due to thought of lethal shocks), solution is immediate debriefing (though it did not necessarily happen!)

Stanford Prison Study

participants were either prisoners or guards, in order to understand how social roles affect behavior

Ethics of Stanford Prison

welfare of the participants were in danger, as guards became aggressive and took away rights and gave punishment; study had to be terminated early

Requirements for Research: anticipate

what might happen

Requirements for Research: react

to what is happening

Requirements for Research: reflect

on what happened and how to prevent future problems

Institutional Review Board (IRB)

necessary in every university, makes sure all research follows ethical guidelines and follows the belmont report principles

Active Deception

being deliberately untruthful (i.e. a cover story)

Passive Deception

omitting key information (i.e. confederates)

Anonymity

participant’s identity is unknown to all (stricter standard)

Confidentiality

participant’s identity is protected from those outside the study but known by those in the study (less strict standard)

Variable

any value or characteristic that can change or vary from on person/situation to another; must be observable and replicable

Observable

variable can be directly/indirectly measured

Replicable

can be consistently observed more than once

Construct

conceptual variable that cannot be directly observed but is known to exist (i.e. children’s learning)

External factor of a construct

observable behavior/event that is presumed to reflect the construct itself (i.e. multiplication tables sheet)

Operational definition of a construct

part of external factor that we observe and measure (i.e. correct number of answers on sheet)

Continuous Variable

measured at any place in a number line (as many decimals needed) (i.e. SES by salary)

Discrete Variable

measured in whole units or categories (i.e. SES by lower/middle/upper)

Quantitative Variable

measured as a numeric value, can be either continuous or discrete

Qualitative Variable

measured in categories or classes, can ONLY be discrete

Nominal Scale

number is designed to represent something (i.e. 1 = male, 2 = female)

Ordinal Scale

numbers represent an order or rank, does not specify space between the ranks

Interval Scale

distributed in equal units with no true zero

Ratio Scale

similar to interval scale but DOES have a true zero

Reliability

consistency, stability, or repeatability of measures of observation

test-retest reliability

extent to which a measure is consistent/stable at two points in time

internal consistency

reflects the extent to which multiple items are given the same picture of the behavior or event being measured

Cronbach’s α

stat that “splits” all items for a measure every possible way and computes a correlational value for them all

interrater reliability (IRR)/interobserver reliability

extent to which two or more raters of the same behavior/event are in agreement with what they observed

cohens κ

stat that gives an estimate of the consistency in ratings of two or more raters

validity

extent to which a measurement for a variable/construct measures what it is intended to measure

Construct Validity

extent to which the operational definition is actually measuring the variable/construct

Content Validity

extent to which items of a measure adequately represent all features of construct

Participant Reactivity

reaction/response participants give when watched

Participant Expectancy

behaves in belief of consistency with study

Evaluation Apprehension

conceals personal info even when anonymous

Participant Reluctance

reaction/response contradicting study intent

Expectancy Effects

researcher’s expectations regarding how participants should behave or are capable of doing

Experimenter Bias

behavior of researcher intentionally or unintentionally affects the outcome of the study

Sensitivity

extent to which an aspect of a study is strong enough

range effect

limitation of range of data measured in which scores are clustered to one extreme

floor effect

task is too hard and all participants do poorly

ceiling effect

task is too easy and all participants do well

error

any influence in response of a participant that can cause variability in their response

Between-Subjects Design

comparing two different groups

Within-Subjects Design

comparing two treatments given to the same group

target population

large/and cannot sample directly

accessible population

smaller/can sample directly

representative sample

sample that resembles target populationn

probability sampling

direct from the target population (used only with animals)

nonprobability sampling

sampling from accessible population (used when testing on people!)

convenience sampling

participants are selected based on convinience

quota sampling

subjects are selected based on criteria in target population

simple quota sampling

little is known about characteristics of target population, so numbers are equal (i.e. equal men and women participants)

proportionate quota sampling

characteristics of target population are known and participants are proportionately represented (i.e. 65% women 35% men)

sampling error

difference between the sample and the population

standard error of the mean

distance that the sample mean values can deviate from the population (numeric measure of sampling error) (0 to infinity)

how to reduce error

increase sample size

sampling/selection bias

sampling procedures that favor certain individuals or groups over others

nonresponse bias

sampling favors participants that request to participate in a research study, but misses participants that choose not to respond

participant pools

college requirement that students must participate in research in order to increase participant numbers