LEARNING GOAL 1.2: DESIGN AN EFFECTIVE PSYCHOLOGICAL EXPERIMENT THAT ACCOUNTS FOR BIAS, RELIABILITY,...

12
LEARNING GOAL 1.2: DESIGN AN EFFECTIVE PSYCHOLOGICAL EXPERIMENT THAT ACCOUNTS FOR BIAS, RELIABILITY, AND VALIDITY Experimental Design

Transcript of LEARNING GOAL 1.2: DESIGN AN EFFECTIVE PSYCHOLOGICAL EXPERIMENT THAT ACCOUNTS FOR BIAS, RELIABILITY,...

LEARNING GOAL 1.2: DESIGN AN EFFECTIVE PSYCHOLOGICAL

EXPERIMENT THAT ACCOUNTS FOR BIAS, RELIABILITY, AND VALIDITY

Experimental Design

Variables

Independent variable: the condition that you change to see if it will have an effect

Dependent variable: the result you measure to look for a change

Experimental research questions often take the form of “How does (IV) affect (DV)?”

Confounding variables: other things that are uncontrolled and could distort the relationship between your IV and DV (these abound in psychology!)

Identifying Variables

You want to see if drinking soda affects dancing ability. What is your IV? Your DV? Your confounds?

You want to test whether sweet foods influence mood. What is your IV? Your DV? Your confounds?

You want to see if people perform differently on tests if they’re reminded of stereotypes beforehand. What is your IV? What is your DV? What are confounds?

Conditions

Experimental: receives the treatment (the manipulated IV)

Control: doesn’t receive the treatmentMay have more than one of either type of

condition

Designing Control Conditions

You want to see if drinking soda affects dancing ability. What should your experimental condition(s) be? What should your control condition(s) be?

You want to test whether sweet foods influence mood. What is your experimental condition(s)? Control condition(s)?

You want to see if people perform differently on tests if they’re reminded of stereotypes beforehand. What should your experimental and control groups do?

Three Sources of Error

Observer ErrorParticipant ErrorAdministrative Error

Observer Error

Example: Confirmation bias (noticing only what supports

his/her theory)How to prevent it:

Random assignment: randomly choose which participant will be assigned to which condition

Double-blind procedure: neither the participant nor the experimenter knows which group the participant is in; may be aided by use of a placebo: an inactive treatment that looks similar to the experimental treatment

Participant Error

Examples: Demand characteristics (trying to give “good”

data) Social desirability bias (trying to “look good”)

How to prevent it: Random assignment Double-blind procedure and placebos Ensuring the participant doesn’t feel watched or

judged

Administrative Error

Example: Variations in how the study is performed

How to prevent it: Double-blind procedure and placebos Strict scripting and clearly defined protocols

Designing Measures

We want our measurements of our independent and dependent variables to be…

ReliableValid

Test Reliability

A measurement is reliable if it gets consistent results

Test-retest reliability: a participant who completes the task multiple times keeps giving pretty similar results

Inter-rater reliability: two evaluators would both score the results the same way

Test Validity

A measurement is valid if it actually measures what it’s supposed to measure

Say you designed a test to measure intelligence based on shoe size. Such a test would be reliable – shoe sizes follow a pretty universal standard – but it could in no way predict intelligence, so it wouldn’t be valid.