Week 8 Flashcards
within group design
compare conditions in same group of participants
between group design
compare across different groups of participants
independent measures
different people in each condition
issues with independent measures
participant variation - individual differences
need to be controlled for e.g. match participants of characteristics
but this is time consuming and we cannot control for all characteristics
repeated measures
same group of people in all conditions
issues with repeated measures
order effects - one condition impacts the behaviour of the other conditions
boredom effects
guessing the purpose of the study - demand characteristics
washout period
period between conditions
removes order effects
when would we use between group design?
distinctive groups e.g gender
quick study
when would we use within subject design?
participant variation
effects of different conditions on behaviour
counterbalancing
avoids order effects
systematically varying the order to remove systematic bias
half participants complete condition a -> b
half participants complete condition b -> a
problems with counterbalancing
more conditions make conterbalancing complicated
latin square counterbalancing
incomplete counterbalancing
randomising the order of conditions
control conditions
act as a comparison
absence of manipulation
e.g. in drug trails, the control is known as the placebo
parametric tests
independent t test
make assumptions about the data shape, if our data doesn’t meet these assumptions then our p value will be misleading
increases the risk of type 1 or type 2 error
non-parametric test
mann whitney u
no assumptions about data shape, applied to any data set
reduces the chance of detecting a true difference
assumptions of a independent t test
data is normally distributed
data shows homogeneity in variance
data measured on interval or ratio scale
homogeneity of variance
dispersion is similar in both groups
if assumption is wrong then stats tests and the p value will be bias
levene test
test for equality of variances
tests the null hypothesis are the same
if the test is significant, then homogeneity of variance Is violated
parametric test shouldn’t be used
independent t test
examines difference in sample means in relation to the standard deviation
underestimating the p value increases the chance of type 1 error
overestimating the p value increases the risk of type 2 error
nominal data
observations placed into categories
categories cannot be ranked in any way
e.g. political party
ordinal data
can be ranked/ordered but there isn’t even spacing
e.g. postions in a race
interval/ratio data
data can be ranked and evenly spaced e.g. weight
difference between interval and ratio data?
ratio data has a true zero point
further up in the hierarchy
more options for analysing data
interval data for example, can be converted into ordinal or nominal but not the other way around
independent t test
differences between the means, in relation to the variability
looks at difference between the groups and differences within the groups
null hypothesis of no differences, t forms dispersion with known shape and can map the positions of p values
mann whitney u test
compares values of each score in a group with each score in other group
how often is group 1s score later than that of group 2s
null hypothesis U forms a distribution to allow a map of p value