Midterm 2 Flashcards
situational independent variable
manipulates situation, example: relaxation or no relaxation
task independent variable
manipulates task, example: solving anagrams or puzzles
instructional independent variable
manipulates instructions, example: solving with time limit or no time limit
statistical conclusion validity
proper analyses and conclusions, respect test assumptions, right type of measures
construct validity
well operationalized independent and dependent variable
internal validity
confound free
external validity
how well can results be generalized to other populations, other environments (ecological validity), and other times
most important type of validity
construct and internal validity
pre and post studies, what could cause changes other than IV?
1) history
2) maturation
3) regression to the mean
4) attrition
subject selection effect
subjects might select themselves into certain groups without random assmt. Monkey ulcer study
between subjects study necessary when…
1) studying subject variable
2) when subjects have to be naive
which random assmt method to use with between group study
- )block randomization-assign participants randomly to all conditions in the block and repeat until all participants have been assigned
- )matching-use when small number of ppl
when to use matching
small number of ppl
problems with within-subjects design
1) order could influence results: progressive effects, sequence effects
2) performance improves over time: practice effects
3) fatigue, performance decreases
complete counterbalancing
way to avoid sequence effects, all possible conditions. For example, 3 conditions=3! within-subjects
partial counterbalancing
random sample of all possible combinations. Within-subjects
block randomization
totally random within block. Within-subject and between-subject
reverse counterbalancing
a-b-c-d-d-c-b-a. Within-subject
ceiling effect
average scores for the group are so high that no difference can be determined between conditions
floor effect
all scores are extremely low
examples of ways to lose internal validity
- take place over long period over time (attrition)
- subject groups non-equivalent (subject selection effect)
- no control group
Hawthorne effect
When behavior is affected by the knowledge that one is in an experiment
cohort effects
groups differ in terms of environment in which they were aged, not just age
demand characteristics
aspects of the study that reveal hypothesis being tested
- might be a reason to use between subject design
- can use manipulation checks
independent groups design
b/w groups, manipulated variable, random assignment
matched groups design
b/w groups, matched assignment, manipulated variable
ex post facto/nonequivalent groups design
subject variable, b/w groups, deliberate non-equivalence
within-subject if subject is tested once. How to randomize?
full or partial counterbalancing AB or BA
within-subject if subject is tested more than once. How to randomize?
reverse/block counterbalancing ABBA
placebo control group
subjects think they are being treated
waiting list control group
receive treatment after study is over
yoked control group
control group yoked to experimental subject
ANOVA: when to use?
more than two-levels in design
types of ANOVA
one-way:one independent variable
two way: two independent variables
what do you need to calculate ANOVA?
mean and number of subjects in each level
F ratio
variance between groups/variance within groups
Variance between groups
sum of squares between groups/degree of freedom between groups
Variance within groups
sum of squares within groups/degree of freedom within groups
reinforcement delay
between groups variance
error
within groups variance
multilevel design advantages for between-subject designs
1) can reveal nonlinear results
2) can rule out alternative explanations
multilevel design advantages for within-subject designs
1) can reveal nonlinear results
2) all counterbalancing options available
when is t-test used
single factor, two level design
t-test for independent groups
manipulated IV, subject IV
t-test for dependent groups
matched groups, within-subjects design
t-test formula
(mean x1-mean x2)/standard error of difference
df formula
n1+n2-2
Cohen’s d, calculates effect size
difference in mean/standard deviation for both groups combined
difference in mean/s
