Hypothesis Testing:
alpha error, type I error, false positive -->
Probability of accepting the experimental hypothesis when it is not true.
beta error, type II error, false negative -->
Probability of rejecting the experimental hypothesis when it is in fact true.
Hexp:
Mean dance score for high alcohol group is greater than mean dance score for low alcohol group.
H0:
Mean dance score for high alcohol group is equal to mean dance score for low alcohol group. (There is no difference in the two groups.)
Qualities of true experiments:
confound: uncontrolled extraneous variable that could provide an alternative explanation
Social desirability bias - when people respond by trying to put themselves in a favorable light
UNOBTRUSIVE and ARCHIVAL MEASURES:
physical trace studies - using physical remnants, fragments, products as evidence of past behavior
erosion (wearing down)
accretion (building up)
archival study - examines already existing records (the experimenter is not collecting new data, but examining data already gathered by someone else).
cross sectional design - one "slice" of time
successive independent samples - different groups measured at different times.
longitudinal design - same group followed over time.
cohort - a group of people exposed to the same set of circumstances (especially a group born at the same time).
cohort effect - effect that is due to the circumstances a group of people are exposed to (and often not the variables of interest).
Ex: Vietnam veterans may score differently on a "startle reaction" test because of combat experiences.
OR
People who lived through the "energy crisis" in the 1970's may be more concerned about conserving energy.
respondant [subject] mortality - the loss of subjects in a study because they drop out, disappear, or die.
Between subjects design--
different groups of subjects serve in the different conditions of the study
Within subjects design--
the same group of subjects experiences the different conditions in the study
Within-subjects advantages:
1) Fewer subjects
2) Less admininstrative work
3) Reduces unwanted error variance between subjects
Ways to deal with practice effects
1) Every possible order
(complete counterbalancing)
2) Rotate order
3) Latin square
(partial counterbalancing)
Ex. of Latin square (subjects exposed to 4 different kinds of music):
cheerful, sad, romantic, patriotic sad, patriotic, cheerful, romantic patriotic, romantic, sad, cheerful romantic, cheerful, patriotic, sad
Repeating conditions in within subjects designs:
1) ABBA (reverse counterbalancing)
ABBA order--When subjects experience conditions more than once, they first experience the conditions in one order, and then the reverse order.
Ex:
Diet Coke
Diet Pepsi
Diet Safeway Select Cola
Diet Safeway Select Cola
Diet Pepsi
Diet Coke
2) block randomization
Block Randomization -- counterbalancing technique to insure that when subjects are tested in each condition more than once, they experience each condition once before experiencing a condition again.
Block randomization is also used in BETWEEN subjects designs--
assures that one subject experiences each condition before a second subject is run again in that condition.
Assigning subjects to conditions in between subjects designs:
- random assignment
- block randomization
- matched groups