11972845
Research
Methods
Week 3
- Data
Types
- Interval - equal increments but no
true 0 - e.g dress size, size 0 doesn't
indicate an absence of size, 0
degrees does not indicate absence
of temperature, 0 on exam doesn't
indicate absence of knowledge
- Ratio - does have a true 0.
0cm does indicate absence
of length, 0g does indicate
absence of weight
- Nominal - category
membership -
gender, religion etc
- Ordinal - ranked or ordered -
race positions, attractiveness,
strongly agree, agree, disagree
etc
- Hierarchy Ratio Interval
Ordinal Nominal
- Interval - equal increments but no
true 0 - e.g dress size, size 0 doesn't
indicate an absence of size, 0
degrees does not indicate absence
of temperature, 0 on exam doesn't
indicate absence of knowledge
- Between Subjects
Designs
- How do we ensure differences in dependent
variable result from independent variable
rather than something else like age, driving
experience etc
- Can't eliminate these
effects but can
minimise them by
spreading their
influence across
different levels of the
IV
- Random Allocation -
ensures each participant
is equally likely to be
assigned to any IV level -
distributes the
occurrence of potential
moderating variables
equally among
experimental conditions,
prevents experimenter
bias, enables use of stats
tests to determine causal
relationships between
variables
- Random Allocation -
ensures each participant
is equally likely to be
assigned to any IV level -
distributes the
occurrence of potential
moderating variables
equally among
experimental conditions,
prevents experimenter
bias, enables use of stats
tests to determine causal
relationships between
variables
- Can't eliminate these
effects but can
minimise them by
spreading their
influence across
different levels of the
IV
- How do we ensure differences in dependent
variable result from independent variable
rather than something else like age, driving
experience etc
- Within Subjects Design
- Potentially moderating
characteristics kept
equal across all levels,
but has order effects
- Once participants have been
exposed to one level of IV
theres no way to return them
to their original state
- Counterbalancing - split
group in half, have half
do AB and other half do
BA - order effects will still
influence, but influence
will be equally spread
across IV
- Counterbalancing - split
group in half, have half
do AB and other half do
BA - order effects will still
influence, but influence
will be equally spread
across IV
- Once participants have been
exposed to one level of IV
theres no way to return them
to their original state
- Potentially moderating
characteristics kept
equal across all levels,
but has order effects
- Factorial Designs
- Experimental designs with
2 or more IVs - allows us to
ask what effect does IV1
have on DV, what effect
does IV2 have, and what
effect does the interaction
between IV1 and IV2 have
- Example - effects of alcohol
consumption and work shift
pattern on work productivity
- Example - effects of alcohol
consumption and work shift
pattern on work productivity
- Fully independent
factorial design
(between subjects)
- Each participant
takes part in only
1 level of 1 of the
IVs
- Each participant
takes part in only
1 level of 1 of the
IVs
- Fully repeated
measures
factorial design
(within subjects)
- Each participant
takes part in all
experimental
conditions (all
levels of the IV)
- Each participant
takes part in all
experimental
conditions (all
levels of the IV)
- Factorial Mixed design
- Always contain at least 1
or more within subjects
IV, and one or more
between subjects IV
- Each participant takes
part in all levels of within
subjects, but just one
level of between subjects
- Each participant takes
part in all levels of within
subjects, but just one
level of between subjects
- Always contain at least 1
or more within subjects
IV, and one or more
between subjects IV
- Experimental designs with
2 or more IVs - allows us to
ask what effect does IV1
have on DV, what effect
does IV2 have, and what
effect does the interaction
between IV1 and IV2 have
- Between subjects
design without
random allocation
(quasi-experimental)
- Quasi-experimental designs -
assignment of Ps is
predetermined - e.g males or
females, alcoholic vs
non-alcoholics
- Have to be cautious
about inferring causality
- Solution - matching - identify potentially
moderating variables and match the
groups based on this - e.g match groups
on IQ, education level etc
- Even better is matched pairs,
but usually impossible to
perfectly match Ps in this way
- Even better is matched pairs,
but usually impossible to
perfectly match Ps in this way
- Solution - matching - identify potentially
moderating variables and match the
groups based on this - e.g match groups
on IQ, education level etc
- Have to be cautious
about inferring causality
- Quasi-experimental designs -
assignment of Ps is
predetermined - e.g males or
females, alcoholic vs
non-alcoholics
- Within subjects
design without
counterbalancing
- Sometimes not possible to counterbalance,
e.g when examining the effectiveness of
mnemonic training on memory performance
- the order in which participants are exposed
to levels of IV is fixed
- A compromise - pretest posttest - split Ps
into 2 groups, treatment group = test ->
manipulation -> test and control group -
test -> no manipulation -> test
- A compromise - pretest posttest - split Ps
into 2 groups, treatment group = test ->
manipulation -> test and control group -
test -> no manipulation -> test
- Sometimes not possible to counterbalance,
e.g when examining the effectiveness of
mnemonic training on memory performance
- the order in which participants are exposed
to levels of IV is fixed
- Developmental terms
- Specific terms when referring to
developmental research - between
subjects = cross sectional, and
within subjects = longitudinal
- Specific terms when referring to
developmental research - between
subjects = cross sectional, and
within subjects = longitudinal
- Measurement error
- Random error
- obscure the
results
- Constant
error - bias
the results
- Random error
- obscure the
results
- Bad variables
- Extraneous -
undesirable variables
that add error to our
experiments
- Confounding variables - extraneous
variables that disproporiantely affect one
level or the IV more than other variables -
introduce threat to internal validity
- Can result in us measuring an effect of the
IV on the DV when it isn't there, or no effect
of the IV on the DV when it is present
- Can result in us measuring an effect of the
IV on the DV when it isn't there, or no effect
of the IV on the DV when it is present
- Confounding variables - extraneous
variables that disproporiantely affect one
level or the IV more than other variables -
introduce threat to internal validity
- Extraneous -
undesirable variables
that add error to our
experiments
- Internal Validity
- Threats to
internal
validity
- Selection - bias resulting from the selection or assignment of
participants to different levels of the IV - results if participants
who are assigned to different IV levels differ systematically
- History -
uncontrolled
events that
take place
between
testing
occasions
- Maturation -
intrinsic changes in
characteristics of
participants
between test
occasions
- Instrumentation
- changes in
measurement
instrument
- Selection - bias resulting from the selection or assignment of
participants to different levels of the IV - results if participants
who are assigned to different IV levels differ systematically
- Threats to
internal
validity
- Reactivity
- Awareness that they are
being observed may alter Ps
behaviour - can threaten
internal validity if Ps are
more influenced by one IV
level than another
- Awareness that they are
being observed may alter Ps
behaviour - can threaten
internal validity if Ps are
more influenced by one IV
level than another
Want to create your own Mind Maps for free with GoConqr? Learn more.