Mann-Whitney U test interpretation in SPSS Video Lecture | SPSS: For Beginners - Data & Analytics

hi everyone Phil from Statistics
mentor.com and hope you're up for this
video on the mann-whitney u test now
think about comparing two groups recall
that we have like either an
independent-samples case or a matched
pairs case once we've decided which
we're looking at going to think go check
is the normally data normally
distributed here we're looking at a case
where we have independent samples and
our data are not normally distributed in
this case won't want to compare two
groups we use the nonparametric
mann-whitney testing so this is what our
data looks like you have two columns one
column denoting the group and one column
with the measurements which will be far
from normal if it's approximately normal
can still use the parametric t-test but
here we're going have it to be really
not normal so our example is of an
advertising agency who which is the test
is so two adverts wants to know which
one works better so two adverts and it
places one advert in a city and its
other advert in a different city and at
the time it gauges the responses by
measuring people's responses on a scale
from one to ten all right the data turns
out to be not to be normally distributed
so we kind of use the nonparametric
mann-whitney test now this scale is one
if they think it's a rubbish add to ten
if they think it's a great cut
to run the mann-whitney test we go to
analyze nonparametric tests now there
are two options let's go through both of
both them give us the same results first
we can use the legacy dialog legacy
dialog means that ancient times really
that's where I see means that in
previous versions of SPSS you if you
learnt it through this through a a
really old text book or something as
opposed to the more kind of newer texts
so you can still run using the old
commands so we look for let's do it
first so nonparametric test the legacy
dialogues I'm quite ancient so I prefer
using a legacy where it's variable two
independent samples click on that okay
so we've got ratings this is our
measurement we move it into the text
variable list our city which is coded
one and two it's grouping variable move
it to their question mark comes up just
like for the independent samples t-test
to find the groups coded it one and two
so we just go one and two right just use
value one to whatever value you used to
denote each of the groups press Continue
make sure the box test type is ticked is
mann-whitney u test and click OK now the
null hypothesis for this test watch
turning away
null hypothesis is that there is no
difference in the middle value versus
the alternative there is a difference in
the middle value okay two tables ranks
let's forget what that is saying and
test statistics
here's a man whitney you test statistic
is thirty four point five from that we
obtain the asymptotic p-value two-tailed
test which comes in at point three eight
four remember that p is low we reject
them nor must go so since P is bigger
than not point naught 5 we do not reject
the null so conclude that the middle the
the middle scores for the both cities
are not different we could write there
is no difference in the ratings of the
two advertisements and then open
brackets right the test statistic these
numbers here or LS c n based on a sample
of 10 there and 9 in city two p-values
point four rounded that okay so we do
not reject the null that's the key thing
going up to the ranks here the mean
ranks are looking at let approximately
the same that's why well
city one gets a mean rank of 11 so
that's higher than two city to right so
that's it that's the mann-whitney test