Week Eight - Experimental Design Flashcards
What are the 3 classes of variables?
Behavioural
Stimulus
Subject
What are Behavioural Variables?
Any observable response produced by a subject.
What are Stimulus Variables?
Environmental factors that have actual or potential effects on behavioural variables.
What are Subject Variables?
Characteristics that can be used to classify a subject for research purposes.
What is a Dependent Variable?
Variable that is measured or observed
How can a DV be measured?
Sometimes DV can be measured directly
Physical arousal
Virus antibodies
Sometimes DV cannot be measured directly and is represented by a construct
Pleasure
Love
Pain
What is an Independent Variable?
A variable that is manipulated by researcher and Requires two levels or conditions
When can an IV not be manipulated?
When it is impossible i.e., sex
When it is unethical i.e., drugs
Experiments with multiple IVs are called?
Factorial experiments
What are Extraneous Variables?
Variables other than the IV that can affect the DV Contribute random (unexplained) error (variability) to the data.
What are Control Variables?
Extraneous variables that are held constant to avoid confounding
Example: Learning can be affected by wakefulness
Control time of day study takes place
What is an operational definition?
Procedures or operations that specify how to manipulate or measure a construct.
Applies to both IVs and DVs.
What are the benefits of multiple operational definitions?
Different aspects of concept can be measured
- Most constructs are multidimensional
Convergent validity
- Different operational definitions yield common findings
Allows for multiple ways to design studies
What are the 3 goals of experimental designs?
Maximise (hypothesised) causal effect
Minimise influence of extraneous variables
Control confounding variables
What factors (4) make experimental designs true experiments?
IVs are manipulated
Operational definitions
Potential systematic effects of extraneous variables/noise are deliberately managed
- Sampling and random assignment
Confounding variables
- held constant
Experiments attempt to do what?
Maximise between groups variance
Minimise within-groups variance
Explain between groups variance?
Caused by an IV and is systematic
But can be affected by confounding variables that may act to increase difference between groups so that the IV appears more important than it is
Decrease differences between groups so that the IV appears less important than it is, masking IV effects
Explain within groups variance
Random (error variance) - Variance caused by unknown factors:
Extraneous variables
Individual differences
Measurement error (unreliable instruments)
Isn’t accounted for IV(s)
Explain quasi-experimental designs
Quasi-experimental research involves the manipulation of an independent variable without the random assignment of participants to conditions or orders of conditions.
Unable to manipulate IV
Subject variables (sex; medical condition)
Unable to control all confounding variables
Can have control variables
Same statistical methods as experimental designs
Explain the process of a single group, post test only (one shot case study) and its characteristics
Poor quality study
- No control
- Not an experiment
Major threats to internal validity
- Bias (sampling method)
- Maturation
- History
- Regression to the mean
- Placebo effects
Low internal validity
Identify: Identify a Sample
Introduce: Introduce intervention (IV)
Measure: Measure DV after intervention
Compare: Compare sample with cases casually observed
Assume any difference with already known cases that did not have IV reflects the effect of IV.
Explain the process of a single-group, pretest-posttest experiment and its characteristics
Simplest form of within-subjects design
- No control group
- Not an experiment
Major threats to internal validity
- Bias (sampling method)
- Maturation
- History
- Regression to the mean
- Placebo effects
Modest internal validity
Sample: Identify a Sample Pretest: Measure DV before intervention Apply IV: Intervention/treatment Posttest: Measure DV after intervention Compare: Pretest vs posttest
Why should we use control groups and random assignment?
Control group reduces threats of history, maturation and regression to the mean.
Random assignment reduces selection bias.
Explain what is a two-group posttest-only, control group design?
Simplest form of between-subjects design
Controls most confounding variables
Random assignment with control group
True experimental design
Sample:
- Random for Control Group
- Random for Experimental Group
Apply IV
- Control = none or placebo
- Experimental = intervention/treatment
Posttest
- Measure DV after intervention for both
COMPARE BOTH
Explain two-group, pretest-post test and its characteristics
Simplest form of mixed-design
True experimental design
- Random assignment
Good internal validity
- Controls maturation
- Controls history
Sample: (random) for control & experimental Pretest: Measure DV beforehand Apply IV: Differs for each group Posttest: Measure DV afterwards Calculate effect: Pretest vs posttest COMPARE EFFECT
What are the benefits of within subjects design? (3)
Every participant is involved in every condition
Maximises group equivalence
- Participants are compared with themselves = perfect matching
Allows estimation of variance due to individual differences
- Increases statistical power
- Correlation of scores between conditions is taken into account.
What are the issues for Within-Subjects Design? (3)
Order effects
Carryover effects
- Later conditions are affected by earlier conditions
- E.g., fatigue or practice
Sequence effects
- Condition order matters
- E.g., Condition 3 affected when immediately preceded by condition 2, but not condition 1
How do we minimise order effects in within subjects design? (3)
Randomise condition order for each participant
- True randomisation might mean some orders never occur
Complete counterbalancing
- Easy with small number of conditions
- Three conditions requires six orders
- ABC, ACB, BAC, BCA, CAB, CBA
Partial counterbalancing
- Latin square
- Each condition occurs once in each possible position
- Number of orders = number of conditions
What is the simplest factorial design?
2x2
What is a Between-Subjects Design?
Where each combination of levels of IVs (cells) is composed of different participants. Different participants in all cells.
What is a Within-Subjects Design?
All participants contribute to all combinations of factors. Same participants in all cells.
What is a Mixed design?
One or more groups (between-subjects factor) and one or more conditions (within-subjects factor).
Explain the partitioning effects in factorial designs.
Variance is partitioned into three effects:
Main effect of (Factor A)
Main effect of Factor B)
Interaction effect of (A x B)
- Does variation in the DV due to A depend on B?
What is a main effect?
A main effect (also called a simple effect) is the effect of one independent variable on the dependent variable. It ignores the effects of any other independent variables
Effect of one factor, ignoring the other factor
One main effect for each factor in the design
Two-factor design has two main effects
What is an interaction effect?
An interaction effect is the simultaneous effect of two or more independent variables on at least one dependent variable in which their joint effect is significantly greater (or significantly less) than the sum of the parts.
Interactions test differences of differences
Number of interaction effects depends on number of factors
- A simple, two-factor design has one two-way interaction
- Three-factor design has three two-way interactions, plus one three-way interaction
Chi-square test-of-independence tests what effect?
Interaction
- Test will be significant if there is an (large enough) interaction