PSYC 255 Exam 2

response set

participants answering without thinking

fence sitting

consistent neutral response

reverse-coded items

What helps catch response sets?

social desirability

answering how we think we’re supposed to

observer bias

seeing what you expect to see in a subject

observer effects

the presence of a researcher affects the outcome of a study

inference

making judgments on a population from a sample

census

studying every member of the population (no inference needed)

convenience sample

measuring those who are easiest to contact or who invites themselves

representative sample

the sample is similar to the population

biased sample

the sample is dissimilar to the population

simple random sampling

sample collected completely at random

cluster sampling

randomly select cluster(s) from the population and then include all from the cluster(s)

multistage sampling

randomly select cluster(s) from the population and randomly select individuals from the cluster(s)

stratified random sampling

divide population into subgroups and then randomly select individuals within the subgroups so that their proportions match those of the population

oversampling

over-representing small subsets of the population in the sample and then mathematically adjusting afterward

systematic sampling

starting point is random, but intervals are equal

probability sampling

type of sampling that always contains an element of randomness so every member of population has an equal chance of being selected, leading to good external validity

purposive sampling

when you need a sample with some specific characteristic, you recruit in a nonrandom location

snowball sampling

ask participants to recommend others who might participate

quota sampling

construct the sample so that it reflects subsets in the population using a nonrandom technique (biased version of stratified)

sampling method

What element of the sample affects external validity?

sample size

What element of a sample affects statistical validity?

random assignment

What increases internal validity?

correlation coefficient

statistic used to describe the relationship between two variables

r

symbol for correlation coefficient

positive relationship

variables move in the same direction

negative relationship

variables move in opposite directions

design confound

variable that systematically changes with the independent variable

selection effects

participants at each level of the independent variable differ in some systematic way before independent variable occurs

random assignment

How can you solve selection effects?

independent groups design

experiment in which you randomly assign participants to different groups and compare measures of the dependent variables between the groups

within-groups design

experiment in which each participant experiences all independent variable conditions

concurrent-measures design

within-groups design in which participants are exposed to all levels of independent variable and the dependent variable is measure once (usually they pick a preference between several conditions)

repeated-measures design

within-groups design in which participants are exposed to all levels of independent variable and dependent variable is measured multiple times

order effects

in within-groups designs, being exposed to one condition can change how participants react to another condition

counterbalancing

solution to order effects that involves switching up the order in which participants are exposed to the conditions

maturation threat

changes in behavior that occur spontaneously over time

history threat

changes in behavior that result from some systematic external event that participants experience between pretest and posttest

attrition threat

loss of participants is problematic when a certain type of person is more likely to leave the study

instrumentation threat

scores change because a measuring instrument changes over time or nonequivalent measures were used

testing threat

scores change because the participant has already taken the test

regression threat

extreme scores come closer to the average over time

demand characteristics

participants change their spontaneous behavior to fie what they think the study is about

placebo effect

expectation-based change in behavior

categorical variable

levels are things that are not quantitative, like eye color

quantitative variable

levels are numbers

dichotomous variable

variable with two levels

ordinal scale

categories with rank ordering, doesn’t provide information about distance

interval scale

equal distance between observations, but no true zero (temperature, shoe size, IQ, etc)

true zero

a zero that is a meaningful and possible value

ratio scale

equal distance between observations AND true zero

reliability

consistency, preciseness, or dependability

test-retest reliability

consistency over time

interrater reliability

consistency across 2 or more raters of behavior

internal reliability

consistency among similar items that measure the same construct

validity

truthfulness of a measure

face validity

extent to which a measure appears to measure what it is intended to measure

content validity

extent to which measure captures all parts of the construct it’s intended to measure

criterion validity

valid measures should correlate with behaviors or outcomes that are related to the construct

known-groups paradigm

rather than using a behavior, are there existing groups that can help validate a measure?

convergent validity

scores on a new scale are related to scores on a similar existing scale

divergent validity

scores on a new scale are unrelated to scores on a different existing scale

leading questions

questions that include the desired response in the item

Barnum questions

questions that apply to virtually anyone

double-barreled question

question that asks more than one thing

anchoring

providing participants with quantitative answer choices that influence their thinking

d

symbol for effect size

.2

small effect size

.5

medium effect size

.8

large effect size

quasi-experiment

almost an experiment but some aspect of experimental control is missing

nonequivalent control group design

quasi-experiment where at least one treatment and one control group, but participants have not been randomly assigned

interrupted time-series design

quasi-experiment where participants are measured on a dependent variable before and after the “interruption” caused by an event

nonequivalent control group pretest/posttest design

quasi-experiment where there is at least one treatment group and one control group, but participants have not been randomly assigned and are tested before and after some intervention

limits and theories

2 things a factorial design is really good at testing

factorial design

experiments with more than one independent variable that allow us to answer how “it depends” or “only when”

interaction

when the effect of one independent variable depends on another independent variable

main effect

an overall effect of one IV

parallel

what do the lines look like on a graph of a factorial design with no interaction?

marginal mean

average for one level of an independent variable