Transcript Components of a causal relationship

Components of a causal relationship
•
•
•
•
•
Does a change in X cause a change in Y?
There are 3 components:
1) Co-variation of events
2) Time-order relationship
3) Elimination of alternative causes.
Independent Variable
• The presumed "cause" of a behavioral effect or
change
• Manipulated (varied) by experimenter
• IV has several levels selected by experimenter
• Occurs, or can be "set up" before DV is measured
• "Independent" of what the subject does.
Dependent Variable
• Some measure of behavior that is a measure of
the effect of the IV(cause)
• What is recorded by the experimenter
• The behavior occurs after IV is varied, and DV
measures the behavior
• "Depends" on manipulation of the IV
• DV does not have levels.
Confounding Variable
• Any variable that is a potential cause for the
experimental effect, other than the IV
• Any variable whose values change
systematically across levels of the IV.
Control variable
• Variable whose values remain the same
across levels of the IV (eg, room temp, light
levels, time-of-day, etc).
Random variable
• Variable whose values vary randomly in an
unbiased way across levels of the IV
• Random variables are usually created by the
process of random assignment.
Subject variable
• A personal characteristic (eg, height,
weight, gender, ethnicity, socio-economic
status, etc).
Control group
• The group that receives “zero” or “the
absence of” the IV
• Eg, the placebo group in a drug experiment
• The group that serves as a baseline to
compare with the performances of the
experimental groups.
Experimental groups
• The groups that receive non-zero values of the IV
• Eg, the drug groups in a drug study
• The performances of these groups are compared
with the performance of the control group.
Conceptual Definition
• Definition of a variable at the conceptual or
idea level
• Tends not to be very precise
• Tends to be more general, more vague.
Operational Definition
• Specifies the operations or procedures necessary
to measure the variable
• Very precise
• Not general or vague at all
• Tells how the variable was measured
• There may be many OD’s for a single CD.
ODs and CDs - Example 1
•
•
•
•
Conceptual - Amount of alcohol
Operational - # of beers in 1 hour (0,1,2,3)
Operational - grams of alc./kg body weight
Operational - BAC (mg alc./deciliter blood).
ODs and CDs - Example 2
•
•
•
•
Conceptual - Helping behavior
Operational - # of people who help a “victim”
Operational - duration of helping behavior
Operational - # seconds before helping occurs
(latency).
EXR-intermediate scenarios
Complex designs
•
•
•
•
•
More than one IV
Eg, Left/Right and 1, 5, or 10 spaces fr. center
More efficient than single IV experiments
Gives more information
Allows analysis of main effects and interactions.
Complex designs - terminology
•
•
•
•
•
An IV is called a factor
number of numbers = how many IVs there are
values of numbers = how many levels each IV has
“2 X 2 design” (two IVs, each with 2 levels)
“2 X 3 design” (first IV has 2 levels, second IV has
3 levels)
• “2 X 8 design” (first IV has 2 levels, second IV has
8 levels)
• “2 X 2 X 4 design” (first IV has 2 levels, second IV
has 2 levels, third IV has 4 levels).
Main effects
•
•
•
•
There is one potential main effect for each IV
A 2 X 8 design has two possible main effects
A 2 X 2 X 4 has three possible main effects
A main effect is present if an IV had a significant
effect on the experiment’s outcome (regardless of
the effects of the other IVs).
Interactions
• Please memorize: “An interaction occurs if
the effect of one IV varies depending on the
level of the other IV”
EXR-horn honks and abstracts
Designing experiments
• Two general types of designs
• Between-subjects (between groups or independent
groups) = each group gets one level of the IV
• Within-subjects (within-group or repeated
measures) = each subject gets all levels of the IV
• Equivalency of groups at each level is built-in for
within-subjects and achieved by random
assignment for between-subjects
• Within - more efficient in terms of # of subjects
• Within - zero variability (ind diff) between levels.
Order effects
• Order effects (practice effects) = experiencing
one level affects behavior in another level
• Eg, does content (biology text vs. novel) affect
proofreading speed? Order is Biology-Novel
• Eg, practice, boredom, fatigue
• Order effects cannot occur in between-subjects
and are controlled in within-subjects by
randomization or counterbalancing.
Differential carryover effects
• (carryover effects, differential/asymmetrical
transfer effects)
• The effect of the first level on the second level
differs depending on which comes first
• Effect of B following A ≠ effect of A following B
• Confound is due to which level precedes which.
FIG: Order effects in proofreading
Group 1 (no practice)
Biology
1
(practice)
Novel
2
(no practice)
Novel
1
(practice)
Biology
2
Group 2
FIG: Differential carryover effects in
problem solving
Group 1
Group 2
(no practice)
Neutral
instructions
1
(practice)
Special
instructions
2
(no practice)
Special
instructions
1
(practice)
Neutral
instructions
2
Other considerations
•
•
•
•
Mixed designs (some between, some within)
Small-n designs
Matched groups designs
Demand characteristics = cues that tell subjects
how they should behave (eg, drug studies)
• Blind and double-blind procedures
• Internal and external validity
• Quasi experiments.
(no practice)
(practice)
Group 1
Neutral
instructions
Special
instructions
1
2
(no practice)
(practice)
Special
instructions
Neutral
instructions
Group 2
1
2