EXPERIMENTAL PSYCH EXAM 3

OBSERVER BIAS

researcher’s expectations influence how they record data, leads to inaccurate results

REACTIVITY

inds change behavior when aware they’re being observed

2 EXAMPLES OF BIASED SAMPLES

convience & volunteer

WHY ARE BIASED SAMPLES, BIAS?

may not accurately represent pop

convience = sample is easy access picked

volunteer = may attract certain group of inds

HOW TO IMPROVE BIASED SAMPLES

use random sampling so everyone has equal chance of selection

HOW DOES RANDOM SAMPLING WORK?

inds picked randomly & have equal chance of selection

WHY RANDOM SAMPLING ALLOWS GENERALIZATIONS TO LARGE POP

reduces bias, represents population, provides accurate results

WHEN & HOW TO ENSURE SUBGROUPS HAVE SIMILIAR PORPORTIONS

ensure groups by certain characteristic

how = stratifed sampling

STRATIFIED SAMPLING

divide pop into subgroups w/ common trait

BIVARIATE CORRELATION

statistical measure seeing how strong relationship of 2 variables is

CONSTRUCT VALIDITY

how reliable & valid were the measured variables?

STATISTICAL VALIDITY (effect size & significance)

what does Pearson's r about relationship strength/effect size? 

what if the confidence interval includes zero? what can outliers do?

can restriction of range skew our interpreation?

INTERNAL VALIDITY

make sure effect isn’t influenced by 3^rd external factor

EXTERNAL VALIDITY

trait of sample is related to the outcome variable, limits generalibility

WHY CAN’T CORRELATIONAL STUDY SUPPORT A CASUAL CLAIM?

causal needs experimental control

doesn’t show cause & effect or 3rd variables

CRITERIA FOR TESTING A CAUSAL CLAIM

covariance

temporal precedence

internal validity

controlled experiments

COVARIATION

cause & effect are related (one changes, so does the other)

IV is manipulated to see affects on DV

TEMPORAL PRECEDENCE

cause always comes before effect

3 INTERNAL VALIDITY

relationship of cause & effect isn’t affected by 3rd variable

cause is from manipulating IV

CORRELATIONAL RESEARCH

study how 2+ variables are related w/o manipulation

probability sampling

inds chosen randomly & equla chance being picked

best way to get representative sample

masked “blind” design

reduce bias & increase reliability of results

participants, research or both dont know key parts

survey design should be

clear & concise

stay neutral

focus on 1 question at a time

case study vs survey

specfic details about ind vs generalized

types of probability sampling

simple random

cluster (subgroups)

stratified

convience sampling

available inds chosen from pop w/ unknown chance of selection

random assignment

used in experiments only inds are random w/ equal chance of selection

random sampling

used in surveys/observations

generalizability

increases external validity