PSYC3042 Final: PSYC3042: Terms P
Only: P
Parametric IV
Variation
As for sensitivity, varying an IV Parametrically = Stepping systematically through
the estimated effective range of the IV.
But, for construct validity, Question is whether the effect varies as expected if
the IV has captured the intended construct.
e.g., Does parametric variation of time pressure (to vary arousal) produce
inverted U changes in performance?
e.g., Do increases in the angle of rotation required to solve a mental rotation task
systematically increase response latencies?
Parallel DV’s
Parallel DVs = Several measures, tapping into the same dependent construct,
included a study.
Examples of multiple measures tapping one dependent construct:
For anxiety: Physiological & behavioural measures.
For pro-social behaviour (Amato): Helping & speaking.
Look at the effect(s) of the IV(s) on each DV:
If the DVs all respond similarly to the IV(s), this supports their
construct validity.
◼ i.e., The DVs mutually validate each other.
Participant Attribute
Threats
Any participant attribute that has the potential to interact with a focal variable
poses a threat to generality.
Such an element may arise because of:
(a) Convenience Sampling:
For external validity, a representative sample is required.
But:
Studies often report convenience sample.
e.g., Quota sampling: Recruit target number for each sub-group in a population
(non-random recruitment).
e.g., 100 males and females, 20-30 years in age.
Non-random recruitment/sampling is not sufficient to accurately assess the
views, abilities, behaviour, reactions, etc., of the population as a whole.
SOLUTION:
If a sampling frame is available (i.e., a list of people in the study’s target
population, e.g., depressed people; or the general population in census data),
options:
Simple random sample.
Proportionate stratified random sample.
◼ i.e., Randomly sample from subgroups (e.g., SES, etc.)
to match population proportions.
◼ If a sampling frame is unavailable:
Cluster sampling.
i.e., randomly sample from clusters of the target population
(e.g., in schools, suburbs, etc.).
Options:
◼ Include all members of a cluster (e.g., all students).
◼ or sample randomly from each cluster.
(b) Psychology Student Samples:
Common type of convenience sample.
Central Issue: Do results obtained in the lab with 1st year students apply to the
broader community?
Non-problematic student samples.
e.g., In perception and cognition experiments, uni students may perform better
overall.
But qualitatively, humans approach many tasks in the same way (especially basic
processes in perception, attention, memory, conditioning, psychophysiology,
etc..)
Problematic student samples.
e.g., In studies of attitudes, personality, organisational psychology, etc., uni
students are not always a useful model of people in the community.
e.g., Sears (1986): US college students, compared with older adults:
have less crystallised attitudes.
have a less-formulated sense of self.
have stronger cognitive skills.
are more compliant with authority.
have more unstable peer-group relationships.
Related Issue: Research with uni students affects social psychology’s view of
human nature.
e.g., it exaggerates the inconsistency between behaviour and attitudes, the
impact of social influence, etc..
e.g., it under-estimates the importance of some determinants of behaviour, such
as:
◼ stable personality dispositions.
◼ material self-interest.
◼ group norms.
(c) Volunteers:
Student or community volunteers bring special problems.
Their results may not generalise, because volunteers may exceed the norm in:
◼ motivation.
◼ commitment.
◼ expectations of successful outcomes, etc..
(d) Attrition:
If there is substantial attrition from a study:
It is not likely to confound the study if attrition is similar over conditions.
But it may severely limit generality.
e.g., If a program or task is too arduous, distressing, boring, etc. (e.g., running up
Mount Coot-tha every day for six months):
The results for “stayers” are limited in their practical application.
Participant
Expectancy
Participant expectancy effects may take several forms:
Demand Compliance.
Participant Resistance.
The Hawthorne Effect.
Positive Self-Presentation.
See each in their respective boxes.
SOLUTION:
As above: Use disguises.
If possible, use a placebo-control.
If the experimental group improves more than the placebo
group, the treatment effect is not due entirely to participant
expectancy.
Ensure participants are blind to the hypotheses & conditions.
Double-blind control = both the participants and the
experimenter are blind to the hypotheses & conditions.
Document Summary
As for sensitivity, varying an iv parametrically = stepping systematically through the estimated effective range of the iv. Parallel dvs = several measures, tapping into the same dependent construct, included a study. Examples of multiple measures tapping one dependent construct: For pro-social behaviour (amato): helping & speaking. Look at the effect(s) of the iv(s) on each dv: If the dvs all respond similarly to the iv(s), this supports their construct validity. i. e. , the dvs mutually validate each other. Any participant attribute that has the potential to interact with a focal variable poses a threat to generality. Such an element may arise because of: (a) convenience sampling: For external validity, a representative sample is required. Studies often report convenience sample. e. g. , quota sampling: recruit target number for each sub-group in a population (non-random recruitment). e. g. , 100 males and females, 20-30 years in age.
