26 terms

# Comparison of dependent & independent samples designs

###### PLAY
Dependent samples design can have problem of
rank order effects or carry over effects. Counterbalancing may reduce impact of rank order effects, but not carry over effects.
provided scores are positively correlated between repeated measures, dependent samples design leads to
less error variance than independent samples design (because participant variables are held constant rather than allowed to vary randomly)
Standard error of M1 - M2 when samples are dependent
is smaller than when samples are independent
For independent samples, M1 and M2 values will be uncorrelated
because, on any replication, different participants contribute to M1 and M2
For dependent samples, M1 and M2 values will be correlated
because, on any replication, the same participants contribute to both M1 and M2
For independent samples, correlation between scores in Condition 1 and Condition 2
should be zero (necessarily so)
For dependent samples, correlation between scores in Condition 1 and Condition 2
should be positive.
Carry-over effects
effects of one condition carry over to the next
eg perception expt - four conditions, A B C D
C = flashbulb, then C will mask effects of conditions that follow it.
eg order 1: A C D B - effect of C will carry over to condition D (and maybe B)
order 2: C B A D - effect of C will carry over to condition B (and maybe A , D)
not solved by counter-balancing
Counterbalancing
e.g. two conditions A and B
half participants do A then B; half participants do B then A
e.g. three conditions, A B C
one third do ABC; one third do BCA; one third do CAB
random permutations
each participant does a random order of conditions
Negative rank order effects
fatigue, boredom
Positive rank order effects
practice, learning, reduction in nervousness or anxiety
Problem of rank order effects can be solved by
counter-balancing
Rank order effects
extraneous influences on DV can arise when multiple conditions are presented, where conditions presented earlier may be responded to differently than conditions presented later.
random allocation refers
to control of extraneous variables, increases internal validity
Measuring same participants (repeated measures design) or matched pairs produces a
dependent samples design
Measuring same participants (repeated measures design) or matched pairs produces a dependent samples design Benefits:
Ensures that distribution(s) of scores on EVs related to participants are held constant from one condition to next. This can greatly reduce the standard error for the difference between conditions
random sampling
refers to how participants are sampled from the population; and ensures that the sample is representative of the population; hence results can be generalised
increases external validity
random sampling increases
external validity
random allocation increases
internal validity
random allocation is used to convert possible systematic errors
into random errors
does not equate groups; rather random allocation
distributes EVs impartially between groups
Independent and dependent samples designs provide two different ways of controlling for EVs related to participants
individual difference variables such as intelligence, experience
random allocation of participants to groups
(independent samples design)
measuring the same participants
(or matched pairs) across all conditions (dependent samples design)
Random allocation of participants to levels of IV
provides some control over all potential participant EVs