exam 3 Flashcards
Criteria for cause
- Covariance - looking at changes in DV after manipulating IV
- Temporal precedence - manipulation of IV comes before change in DV
- Internal validity - manipulating IV while holding all other factors constant
Characteristics of experiments
- Empirical/objective approach - gathering information through the use of your five senses during systematic hypothesis testing
- Manipulation of variables - value of interest is manipulated and outcome is recorded
- Keeping other factors constant - consistent variation in all conditions
- Determining causal relationships
Independent variable MUST have…
-One IV in each experiment
-2 levels
-Can have more than one IV
-Can be discrete or continuous
Between subjects independent variable
different groups of subjects, different subjects in each level
Within subjects independent variable
Same subjects experience each level of IV
Types of manipulations
-Environmental - changing social or physical environment
-Instructional - changing instruction given to one group over another
-Invasive - create physical changes in participants
How to pick levels of IV
- trial and error - look at past research and choose ones that seem appropriate and hope for the best
- pilot study - pick levels that seem appropriate and try them out on a handful of participants
Manipulation check
institute measures within the study to determine if you actually manipulated what you think you did
Dependent variables must be
- reliable
- valid
- sensitive
- no ceiling/floor effects
Simple random assignment
every participant has an equal chance of being in each group. on average, groups should end up being equal in most characteristics
matched random assignment
matching people on a characteristic that is relevant to the experiment. requires a pretest and used when there is a possibility that random assignment will still cause inequalities between groups
advantages of within subject designs
-fewer participants are needed to detect differences
-more powerful
disadvantages of within-subjects designs
-order effects - practice effects- people improve b/c they have already done it before
-carryover effects - effects of condition A are still lingering when condition B begins
counterbalancing
presenting levels of the IV in different orders to different participants
threats to internal validity
- misc. design confounds - type of extraneous variable that varies systematically with the independent variable
- biased assignment to conditions
- differential attrition - people in one condition drop out at higher rate, no longer have random assignment
- pretest sensitization
- history - behavior is affected by a person’s past experiences
- maturation - developmental changes may take place and subjects behavior changes
one way experimental design
- contains only one IV
- simplest one way design: one IV two levels
- can be within or between subjects
factorial designs
-contains 2 or more IV
- designs can be between subjects, within subjects, or mixed
inferential statistics
we can specify the probability that differences are due to error variance but cannot be 100%
H0 (null hypothesis)
group/conditions will NOT differ. the independent variable will NOT have impact on behavior
Ha (alternative hypothesis)
group condition WILL differ. the independent variable WILL have impact on behavior
rejecting the null hypothesis
high probability that the null hypothesis is false, saying group conditions will differ
fail to reject null
low probability that null is false. null is correct.
Type I error
rejecting null when it is true. you should have failed to reject it
Type II error
Failing to reject the null when it is false. Should have rejected
Alpha level (p value)
probability of making type I error. the smaller the alpha, the smaller chance of type I error
Beta level
probability of making type II error
Power
ability to detect true differences between groups. ability to correctly reject the null. power is opposite of beta
Power analysis
tells you how many subjects you need in your study, given your alpha level and desired amount of power
T
actual difference between group means/expected difference between means if only due to error
Unpaired t test
one way with two levels; IV is between subjects and using random assignment
Paired T test
One way with two levels; IV is within or between using matched assignment
Equation for significant results
t(df) = t value, p<,05
Equation for nonsignificant results
t(df) = t value, p >.05
test when you have one IV two levels
t test (paired or unpaired)
test when you have one IV > 2 levels
one way ANOVA
test when you have more than one IV
factorial ANOVA
ANOVA evaluates..
how much systematic variance there is in your study, how much error variance there is in your study
systematic variance
variability due to your manipulations
between groups
error variance
variability due to individual differences and mistakes
within groups
Post-hoc tests
used to pinpoint exactly which conditions differ from each other
Larger F value indicates
differences likely due to IV
Smaller F value indicates
differences are likely due to error
F formula
F = MSbg/MSwg
