Exam 3 Content - Threats to Internal Validity

Parson's Intro to Experimental Psych

Between-Subject Design

different people are assigned to different groups

within-subjects design

same people assigned to different groups each time

Stroop effect

a psychological phenomenon that occurs when it's more difficult to name a color when it's used to spell the name of a different color

Interaction

when the rate of change is the same, there is interaction.

• when the rate of change is not the same, there is no interaction.

3 basic sources of confound

1. experimenter can affect the design

2. participants can affect the design

3. design can affect the design

Experimenter effects design

1. experimenter attributes

2. experimenter expectations

experimenter attributes

age, gender, and occupation of the experimenter can influence how participants react/answer to the study

experimenter attributes solution

neutralization of attributes or elimination of attribution all together.

experimenter expectance

unintentionally influence the study because we know what we're going to find because we're doing the measuring

experimenter expectance solution

eliminate unintentional influence by using double-blind if possible

2 demand characterists

participant sophistication & positive self-presentation

Participation Sophistication

when the participant is sophisticated enough to know what we are trying to do and their results end up as placebo effect or them messing with us

Participation Sophistication solution

single-blind studies

Informed Consent's effect on participant sophistication

want to give enough info to where the participants know what they are getting into, but not too much to where we are telling them the hypotheses

Mock Jury: subject sophistication example

ineffective because people know it's not a real trial, so, in turn, they take more risks than they naturally wouldn't have

Positive Self-Presentation

people want to be preserved as good, so they will go as far to lie to preserve good self-presentation

Positive Self-Presentation solution

self-consensus: need to word questions so participants answers are not wide spread (ie: what you know others do).

History Effect

when something could have occurred between time 1 and time 2 that affected the participants and then the data.

History Effect solution

1. interview participant(s) before second time

2. run a control group parallel to the treatment group

maturation

internal, subtle changes; people get better or worse at things depending on maturation as time goes on (the carrying-over effect is similar)

maturation solution

counterbalance

carry-over effect

something you do in one situation carries over to the next - people usually get better at the task; pretty inavoidable

counterbalance

switching up the order of tests

Latin square

A formal system of partial counterbalancing ensures that each condition in a within-group design appears in each position at least once.

Latin square/counterbalancing downsides

1. lacks variety

2. pain in ass to run all conditions

between subject issues

1. selection bias

2. loose protocol

3. regression to the mean

4. mortality

selection bias

A polling error in which the sample is not representative of the population being studied, so that some opinions are over- or underrepresented

selection bias: random selection

selection by chance

selection bias: random assignment

different groups by complete randomization (takes more into consideration)

Regression to the mean

the tendency of extreme scores on a variable to be followed by, or associated with, less extreme scores

Regression to the mean solution

do not make big changes due to an outlier, wait until things normalize to see if change is effective.

Mortality

concerned with if there is a bias of people who drop our study

advantages of within-subjects design

1. more powerful (likeliness of big differences; big differences within the same person).

2. fewer people (cost and time effective)

disadvantages of within-subject design

1. subject sophistication

2. balance

3. predetermined IVs

4. between is easier to manage

power

what we hope to find
- 1-B (100% of the time whatever is not beta, is power)

false alarm (alpha)

false positive; type one error

• when we say something is true when it is really false

miss (beta)

false negative; type two error

• when we say something is false but it is really true

null hypothesis

Ho; default, no difference between the groups
- Ho: m1=m2

alternative hypothesis

Ha: alternative, there is a difference between the groups 
- Ha: m1 does not = m2
- Ha: m1>m2

two-sided

non-direction; Ha: m1 does not = m2

one-sided

directional; Ha:m1>m2