T-Tests- Unrelated (Independent samples) and
Related
Unrelated T-tests / 2-Sample test /
Independent t-test
whether two groups of scores have
significantly different means
Between-subjects--> These involve different participants
taking part in different conditions usually an experimental
group and a control group
T-scores--> t=(mean of sample 1)-(mean of sample 2) /
standard error of differences between means
Group X [recalled more words correctly] (mean=__, SD=__) than Group Y
(mean=__, SD= __). An unrelated t-test indicated that this difference was
significant, t(df)=__, p=__ .
Assumes the variance of the two samples are similar
Can be combined to produce an overall estimate of the
spread of the data--> Levene's test
Mann Whitney U-test
data in one/both conditions are not normally distributed (badly skewed
data is in the form of ranks
variances (differences) of the two groups differ
group sizes differ considerably
Participants in Group X were more [accurate] (mean=__, SD=__) than
participants in Group Y (mean=__, SD=__). A Mann Whitney U test indicated
that this difference was statistically [significant] , U=__,N1=__, N2=__, p=__
Badly skewed--> non-parametric equivalent
Levene's test for equality of variance
whether or not the homogeneity (quality of
being the same) of variance
(change/difference) assumption is violated
if the p-value for the Levene's test is
'non-significant,' then the 'variances are
not significantly different'
eg, variances are similar between
the two levels of treatment
Significant--> 'variances not assumed'
Non-significant--> 'variances assumed'
Significant= Equal variances not assured
Non-Significant= Equal variancesassured
Degree of Freedom--> Nsample1 + Nsample2 -2
Related T-tests / Paired Samples T-test
comparing the means of two related samples of
scores to determine whether the two means
differ significantly
Within-subjects--> involves the same participants taking part
in 2 or more conditions of the same experiment
Practice effect- can improve between Condition X and Y.
Fatigue effect- get worse between Condition X and Y
Counterbalance with X-Y, Y-X or
three conditions
Carryover effect- earlier condition affects
performance of a subsequent condition
When tasks differ in difficulty-> change their
behaviour on later tasks
increase time between conditions
T-scores- t=difference in means
/ standard error of the
difference in means
Participants were [faster to respond] in Condition X (mean__, SD=__) than in
Condition Y (mean=__, SD=__). A related t-test indicated that the difference in
[reaction time] was statistically significant, t(df)=__, p=__ .
Wilcoxon Matched Pairs Test
both conditions are not normally distributed
difference can only be ranked in size
variances (differences) of the two conditions are very different
more likely to detect significantly differences
SPSS: T= smaller of the sum of ranks (positive/negative skewness)
Participants [were slower to respond] in Condition X
(mean=__, SD=__) than in Condition Y (mean=__, SD=__).
A Wilcoxon matched-pairs test indicated that the
difference [in reaction times] between the two conditions
was significant. Z=__, N=__, p=__.
Badly skewed--> Non-parametric equivalent
Problems surround using a large number of t-tests in one study
Multiple comparisons and Type 1 error
variances (differences) of the two conditions are very different
more likely to detect significant differences
There was a significant association between line-up type and response, c2 = 5.83, df = 1, N = 70, p =
.016. With sequential designs, participants were more likely to correctly identify the perpetrator and
less likely to not correctly identify him or her. With simultaneous presentations, participants were
less likely to correctly identify the perpetrator and more likely not to correctly identify him/her.