Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
46 Cards in this Set
- Front
- Back
factorial design
|
have experiments with more than one IV; multiply variables to get possible number of conditions.
|
|
main effects
|
average of one variable across the other variables; use one variable and average as if its constant. effect of just one of the IV. Just A
|
|
interaction
|
Combining A and B; the effect of one IV changes in effect to another IV
|
|
two way interaction
|
the effect one IV changes across another IV
|
|
three way interaction
|
pattern between two way interaction changes across the level of a third IV
|
|
within-subjects design(repeated measures)
|
each person participates in all levels of the IV/condition. within a person each person goes through each variable manipulating IV within a group of individuals.
|
|
between-subject design
|
each person participated in only one condition. protect from order or carry over effects
adv: cost efficient and control over participants. each subject is his or her own control. no diff. in subjects dis:participants have to do diff. conditons in specific order. participants can develop diff. attitudes as experiment continues. subjects can experience carry over effect or order effect |
|
carry over effect
disadvantage |
subjects result/effect of one condition is affected by the preceding condition
|
|
order effect
disadvantage |
having participants go through conditions in a diff. order to control for interactions
|
|
quasi-experimental method
|
misisng something that makes them a true experiment; no comparison or control part or lacks randomization
|
|
observation-treatment-observation
|
pretest-post-test; one treatment given before one after; have small group of subjects who you want to apply a new treatment to but dont have a control group ; no background. cant tell if change is made by treatment.
|
|
interrupted-time-series design
|
lots of observations before treatment; have times series of measurment/observations, then interrupt wiht treatment; no control group; can see changes in DV
|
|
non-equivalent control group design
|
have a group of subjects but want a control group so you put a non randomized group together
|
|
matched-groups design
|
creates non equivalent control groups but matched with treatment group with similar variables(variables you can control)
|
|
natural manipulations
|
look at measures but dont manipulate anything; differences in sex, gender, handedness, pathology...
|
|
handedness
|
left handed people die younger. more right handed people than left
|
|
cross sectional method
|
diff. ppl at diff. ages
|
|
longitudinal
|
same ppl at diff. ages
|
|
cross sequential
|
diff. ages, get measures two times. mix of cross sectional and longitudinal
|
|
history and threats to internal validity
|
something happens around the time of the treatment; outdide of the subject in the world
|
|
maturation and threats to internal validity
|
changes in body; how infants change as they get older; change inside subject, effect from inside the subject(change in brain/body)
|
|
regression to the mean and threats to internal validity
|
when you look at a person initially and then look another time, the second measure is closer to the mean than the extreme first measurement( all hitters whether in a slump or doing well at the begining all end up around the same average)
|
|
mortality/subject attrition and threats to internal validity
|
subjects dont come back, lose subjects becasue they didnt eant to or couldnt find them
|
|
individual diff./subject variables and threats to internal validity
|
if treatment didnt work they arent likely to to return
|
|
chi-square test for independence
|
same logic as correlation; when a null hypothesis is true; used to compare observed data with data we would expect to obtain from a more specific hypothesis- null
|
|
contingency table
|
used to record and analyze relationships between 2 variables
|
|
correlational methods
|
correlation doesnt equal causation
|
|
Pearson's product-moment correlation coefficient
|
problem id directionality, confounding variables-intervening variables
|
|
scatter diagrams
|
one extreme value effects the mean and size of the correlation; assume the relationship is linear
|
|
intervening variable
|
when the unused variable affects the outcome
|
|
directionality
|
does A cause B or does B cause A?
|
|
truncated range
|
not looking at entire range the alters the conclusion
ex. SAT scores and college acceptance |
|
nonlinearity
|
relationship between variable is not statistic or directly proprtional to the input, but dynamic and variable
|
|
hypothesis testing
|
approach to answering questions about results
|
|
significance testing
|
how likely it is that the results occured by chance
|
|
null results
|
you cant prove a negative, so the same is true of a null hypothesis. you can say we have no evidence to reject the null hypothesis.
|
|
null hypothesis
|
might be falsified by a specific statistical test
|
|
null hypothesis significance testing
|
through replicating variables, whats the probability of getting the same results through different conditions? if null hypothesis is true how likely are we to observe those same results. how likely is it that changes in the study may be due to error- random variation?
|
|
random variation
|
error
|
|
type one error
|
false positive, leads you to believe there is systematic variance but isnt
|
|
type two error
|
false negative
|
|
eliminate null hypothesis
|
cant prove the null hypothesis; how likely is it that i would observe the same results if the null is true? if probability is small it is unlikely to be true
|
|
ceiling effect
|
all stats near top limits of the DV; making test too easy so everyone gets an A. IV has no room to move up more
|
|
floor effect
|
all stats near bottom of DV; making test too hard os everyone fails. IV has no room to move down further
|
|
null results
|
can reject but never prove null hypothesis
easiest way to get null is small samples with lots of variance |
|
4 ways to increase power of an experiment
|
1. increase effect size; increase diff. between mean distribution, when you increase diff. you gain power.
2. increase alpha to decrease beta you can increase power 3. decrease error and variability by increasing experimental control 4. increase number of participants/population |