-
-
Quantitative variable relationships
- linear: positive, negative, no correlation
- non-linear: curved
-
Quantifying the relationship
- correlation coefficient: pearson's r represents the strength
- interval & ratio scales
- correlations are rarely bigge than .30
- measurement error tends to produce weaker correlations: one way to think about strength is the mean difference
-
Describing relations involving one categorical variable
- groups differ in mean score?
- direction?
- size of the difference?
-
Significance test
If probability is low the relation is real
-
-
One variable cause another
- making infeerences about causality: one variable causes another to occur
- directionality problem
-
3rd variable problem
- 3rd variable cause both variables to be correlated
- very common
- there can be more than one
- indirect & direct
- negative & positive
-
-
DV & IV
- IV has levels
- they both have operational definitions: not a causal statement
- random assignment: different types of people are equally distributed to each condition
- manipulation of the IV: eliminates the 3rd variable problem
-
-
Internal & external validity
- internal: confirm in the experiment
- external: confirm outside the experiment
-
Experimental control
- maximizes internal validity
- sometimes limits external validity
- external validity matters, but without internal validity we dont know what's causing what in either the lab or the real world
-
Increasing external validity in experiment
- replicate with many samples
- when studying abstract variables, ues many concrete instances as feasible
- create a naturalistic atmosphere, both physically and/or psychologically with deceltion
-
Interpreting "null" results
- was your manipulation of your IV powerful enough?
- was your DV sensitive enough to detect differences in the score?: Ceiling effect & floor effect
- did you have enough people in your groups to detect a difference?
- was there a lot of random errorm in the measure of your DV?
-
Units of observation
- smallest level of data on all variables
- each unit if observation can be any defiable unit for which the variable differs from one unit to the next
-
Within-subject manipulation
-
Between subjects
different people are assigned to different conditions (or levels) of the IV
-
Within subjects
(repeated measure) the same people are exposed to all levels or conditions of the IV
-
Benefits of within-subjects design
- no random sampling error; only measurement
- mortality not a confound
- requires half as many subjects
- greater statistical power to eliminate chance as an explanation
-
Weaknesses of within-subject designs
- order effects
- history effects: event ex. 9/11
- maturation effects
- carryover effects: something about the first condition caries over into the second condition
-
Carryover Effects
- fatigue effects: people get tired or bored during the study; performance effects
- practice effects: more engaging; having pratice
- testing effects: familiarity; study
- contamination effects: what happens in the 1st condition spills over and "contaminates" the 2nd
-
Counterbalancing
- order effects becaome confounds, if one condition always comes first
- confound is eliminated if each condition comes in each order
- large order effects can still weaken the observed effect of the IV
- complete: every subject gets every order
- incomplete: different subjects get different orders
-
Factorial Design
- more than one independent variable
- each level of the first IV is paired with each level if the second IV
- ex. 2x2= 4 conditions
- 2 experiments in 1
- matrix table
- interaction: the difference between the IV's
|
|
