EXPERIMENTAL PSYCH TEST #4

simple experiments, threats to validity

control variables help eliminate design confounds

keeping other factors constant so only IV is the effect

 random assignment establishes internal validity

ensuring groups have similiar traits

MATCHING

putting ppl in groups with same traits

explain situations in which matching may be preferred to fully random assignment

small sample size or when a trait could influence the outcome

POSTTEST DESIGN

when do researchers use?

measure effect of IV after treatment

PRETEST/POSTTEST DESIGN

when do researchers use?

measure IV before & after treatment

DESIGN CONFOUNDS

THREAT TO INTERNAL VALIDITY

outside variable changes similiar to IV

makes it hard to see true cause

SELECTION EFFECTS

THREAT TO INTERNAL VALIDITY

groups start with different traits that can lead to bias responses

ORDER EFFECTS

THREAT TO INTERNAL VALIDITY

order of conditions affects response

(within groups)

ARTIFACTS

THREAT TO INTERNAL VALIDITY

unintended distortion or bias caused by outside factors

CONFOUNDS

THREAT TO INTERNAL VALIDITY

Makes it hard to determine true cause of the effect 

consistently linked w/ IV (predictor) or DV (outcome)

HISTORY EFFECTS

external factor(event) affects all/most in group

Might change how they feel or respond to questions 

MATURATION EFFECTS

Natural dev or spontaneous improvement that happens over time, leading to group changes that aren’t caused by the treatment

Ex: disruptive boys calm down as they get used to camp setting

REGRESSION TO THE MEAN

extremely high or low prettiest scores even out when measured again 

(events causing extreme won't be there post)

ATTRITION

ppl drop out, move, die & their scores are dropped 

INSTRUMENTATION EFFECTS

group changes cuz measurement instrument changed 

OBSERVER BIAS

knowingly or not, knowledge of treatment may influence their observations

DEMAND CHARACTERISTICS

participants guess study’s purpose & change behavior to expected direction 

PLACEBO EFFECTS

improve only cuz they believe treatment is working 

TESTING EFFECTS

repeatedly tested on same material, makes participants better from practice instead of treatment

comparison groups (including wait lists) & double-blind studies can help reduce these threats to internal validity

REASONS FOR NULL EFFECT

not enough between-group differences

within-group variability hides group diffs

no actual effect

null effects can be hard to find

inadequate variance between groups

how to identify these problems

weak manipulations

ceiling & floor effects (scores too high/low)

do a manipulation test

how large within-group variance can obscure a between-group difference

hard to see differences cuz within group variance hides effect

causes of within-group variance

individual differences get w/ a repeat measure 

situation noise external distractions (stuff happens)

how to reduce within-group variance

individual differences = change to a within or between groups design, add ppl 

situation noise solution = Control surroundings 

FACTORIAL DESIGN

has more than 1 predicitor (IV)

reasons to use a factorial design

Test multiple hypothesis using same group of ind 

More variables = more complex design & analysis (more possible interactions

EXPERIMENTAL INTERACTION

effect of 1 IV depends on another making it hard to separate their effects

 Describe an interaction as a

“difference in differences”

difference between groups changes at diff levels of another variable

keywords that indicate factorial-design language in a journal article or in the news

INTERNAL VALIDITY THREAT

not enough relevant data to support conclusion

not enough work to rule out alternative explainations

EXTERNAL VALIDITY THREAT

conclusion isnt generalizable to society

CONSTRUCT VALIDITY THREAT

test doesnt measure what it’s supposed to

SELECTION BIAS

inds not picked randomly

doesnt accurately represent group

results = misleading

NONRESPONSIVE BIAS

participants unable/unwilling to participate

limited data

HAWTHORNE EFFECT

ind change behavior when aware theyre being observed

true experiemnts test causal relationships so they must

have an IV

control for confounding variables

rely on random assignment

BETWEEN-SUBJECTS

each subject receive only 1 level of IV

random assignment reduces group diffs

WITHIN-GROUPS

each ind experience all levels of IV

oder & testing effect can be bias so order of levels = counterbalances

(assign diff inds to recieve conditions in diff order)