6: small n designs Flashcards
small n research (idiographic/ morphogenic)
Each subject = separate experiment
Use of within-subject designs
Presented data: individual subjects
Comparisons between subjects
Reliability is assessed by replication (repetition)
where is the small n approach useful?
Assumption of minimal biological or psychological variability
*Much of neuroscience
*Psychophysics
*Cognitive and behavioural neuroscience
*Human clinical neuropsychology/neurology
*Clinical psychology
*Animal learning and cognition (tradition?)
what is the small n approach useful
- Constraints: availability, convenience
- No need for generalization
- Assumption of low biological variability
- Interest in a very small layer of the population
- Need to understand the process in time
- Detailed comparison of subjects
key principles in small n research
Consistent level: magnitude of treatment effects.
Consistent trend: unidirectional changes.
Stability: consistent level or consistent trend.
Search of temporal patterns: time series (analyses).
Search of spatio-temporal patterns: sequential analysis of
behaviour. Serial configuration of events and actions in time.
what is the principle of experimental research
replication
why does small n or single subject research work
Large number of observations per subjects: fluctuations average out.
Factors that can contribute to variability are controlled in a strict manner.
Focus on “powerful” variables with a clear effect: Effect sizes.
problems with small n research
Similar problems as for within-subject designs
*Irreversible changes in the behaviours of the subjects induced by the experimental procedures.
Weak effects (from the IV) on the DV.
Unstable DV’s despite strong experimental control:
*Uncontrolled variables
*Extraneous variables
types of designs
baseline designs
dynamic designs
discrete trial designs
what are baseline designs
Typically what researches refer to when mentioning “single-subject designs”.
types of dynamic designs
Good for moment-by-moment changes in behaviour
what are discrete trial designs
Good for a focus on performance of individual subjects (e.g., psychophysics).
issue to keep in mind with baseline designs
reliability»_space;> would replication produce same (or very similar) results
The behaviour is analysed based on both within and between experimental treatments data.
* No averaging of the data!
what is the baseline phase
Behavioural baseline needs to be established until a
stability criterion is reached.
what is the experimental or intervention phase
expose the subject to each
treatment, until criterion is reached. Then repeat.
ABA(B) designs: baseline - treatment - baseline (- treatment).
This is an intra-subject replication where subjects are their own control(s) (as in group within subject designs)»_space;> internal validity.
reversal strategy in baseline designs
ABA(B) = reversal strategy. You expect to “recover” the original baseline (A) after the treatment (B).
Data from the baseline and treatment conditions are compared for statistically significant differences.
If you run more than one subject (small-n design), you can provide inter-subject replication»_space;> external validity. Typically 3 to 6 subjects.
typical issues with baseline designs
Stability criterion
Uncontrolled variability or variation
Irreversible baselines.
what is the stability criterion
choice is “subjective” or based on previous research / pilot study.
- Removes “transitional” data (unless relevant).
- Fixed time or trials: Give a time-limit to reach crit
what is uncontrolled variability or variation
unstable, drifting baselines. Caused by extraneous variables
what are contrast approaches according to bordens and abbott
“The group approach assumes that if experimental controls fail to reduce uncontrolled variation, then statistical methods should be used to control it. The single-subject approach assumes that if experimental controls fail to reduce uncontrolled variation, then one should endeavour to identify extraneous variables responsible for it and bring them under experimental control.” (Bordens and Abbott, 2008, page 355).
types of replication
exact or direct replication
systemic replication
what is exact or direct replication
part of the single-subject procedures
what is systematic replication
above and beyond that procedure
problems with baselines
drifting baselines
unrecoverable baselines
unequal baselines between subjects
inappropriate baseline levels
what are drifting baselines
slow, systematic changes
what are unrecoverable baselines
reversal does not work due to carryover effects
what are unequal baselines between subjects
individual differences in baseline levels
what are inappropriate baseline levels
Low baseline good unless data has a floor effect, or high baseline good unless data has a ceiling effect.
what are some single subject baseline designs
single factor designs
multifactor designs
multiple-baselines designs
what are single factor designs
Single independent variable. AB (necessary if irreversible changes are present), ABA or ABAB designs
what are multifactor designs
Two or more independent variables. E.g., melatonin + bright light
what are multiple-baselines designs
Several dependent variables. E.g., polygraph.
what are dynamic designs and why are they good
Transitional processes and behaviours are crucial and measured.
Excellent to measure behavioural/process dynamics.
Useful also for continuous variations of the independent variable.
The dependent variable is typically continuous.
These designs are often called “time series” designs.
examples of dynamic designs
Handwriting behaviour and stress/anxiety
* IV: stress condition, then non-stress condition
Galvanic skin response and stress/anxiety
* IV: questions, sequentially presented
what are discrete trial designs
Subjects receive each treatment conditions many times (dozens,
hundreds of times) = trial = datum (one data point).
Extraneous variables are strictly controlled.
Randomized or counterbalanced presentation of treatments (conditions).
Inter-subject variation is often analyzed: comparisons between subjects.
what is sequential analysis
Also called “sequential hypothesis testing”.
The sample size is not fixed in advance: sequential estimation is used.
Data are evaluated as they are collected.
Sampling is stopped in accordance with a pre-defined stopping rule as soon as significant results are observed and/or a quota “N” is attained
Advantages of sequential analysis and when is it useful
Advantage: Potentially economical (subjects, time, money), but also seeking to attain a “n” that could lead to statistical significance.
*Conclusion can be reached earlier than with common hypothesis testing principles.
Often used in clinical trials (biomedical research).
Useful if subject availability is a concern.
Compromise between idiographic and nomothetic research
Procedure: Let’s assume 2 completely randomized (independent group) trials:
If n subjects in each group are available, an interim analysis is conducted on the 2n subjects.
A statistical analysis is performed to compare the two groups, and if the alternative hypothesis is accepted (i.e., the groups are different), the trial is
terminated
what happens if the alternative hypothesis is not accepted
Otherwise, the trial continues for another 2n subjects, with n subjects per group. The statistical analysis is performed again on the 4n subjects. If the alternative hypothesis is accepted, then the trial is terminated. Otherwise, it continues with period evaluation until N of the 2n subjects are available.