SPSS for newbies K or K-1 dummy variables for regression Video Lecture

hey hi everyone this is Phil from
statistics mentor comm in this video I'm
going to be talking about including
dummy variables in a regression now
recall that rule that if you have a
variable with K categories you set up K
minus 1 dummies and put that into your
model with an intercept failure to do
this results in what they call a dummy
variable trap so if you try to put in K
dummies plus an intercept you're going
to fall into what I call a dummy
available trap and all that means is
that you've got case of perfect
multicollinearity meaning that you've
got a perfect linear relationship among
your dummies which is certainly true
because if you know K minus 1 of the
values you can definitely determine the
K value let's look at it in context of
this question I have my DV it's a
standard regression my DV is continuous
it's called levels and I got three
dummies first second third depending on
which treatment they take first type of
treatment second type of treatment a
third type of treatment so here I have
three dump three categories let's see
what happens if I try to include
inception all three dummies in there we
know we should do this because it
results in the dummy available trap so I
take the DV into that box I put in first
second third dummies lay all the dummies
into the independent box let's press ok
right what we find is that SPSS does
generate output but look at what's
happened so our constant which is the
intercept treatment one treatment three
words treatment two gone it's excluded
it because if it didn't exclude one of
the treatments you have the dummy
variable trap which means that you
wouldn't get any output at all so since
you haven't selected one to omit oh you
haven't selected the reference category
it's selected one for you and it's
selected treatment
now let's see what happens if we include
all three treatments by exclude the
intercept so we analyze regression
linear in first second third to exclude
the constant the intercept we go options
then here in this box uncheck include
constant equation okay right what we see
is now it's included all three in so
going back to the root in this case K is
3 with K categories we include all K but
exclude the intercept as we have done
here it works or we include intercept
but only use K minus 1 so we drop one
leaving one out as the baseline or
otherwise known as the reference
category let's compare which one you do
doesn't actually matter in terms of
coefficient interpretation because they
should be consistent the results of
consistent so in this case treatment one
two three all in there saying that
treatment 1 the predicted value of the
DV is 2/5 to treatment two predicted
values 304 3 it's about 305 how does
that how does that cell link up match up
with the other case where we've got K
minus 1 dummies right now here here the
numbers when we have K minus 1 dummies
now when we have dummies the
coefficients interpret represent the
differences between this group and the
one that is the reference category now
the reference category here is 2 because
it's been omitted and a constant is
actually text the value of the reference
category so the predicted value for
treatment 2 is 300 point if we go back
to here for the
will include all intercepts 300 point 4
if we look at the treatment one here
this is saying that compared to
treatment to treatment one level the DV
value is lower by 40 about 48 so in
other words 300 subtract 48 so we're
looking at a figure of about 2 5 2 for
treatment 1 ok treatment 1 2 5 2 from
there you go going back up to table
right treatment 3 this is saying that
compared to treatment to treatment 3 has
a higher DV predicted higher DV by about
5 so in other words you expect treatment
value to be 305 treatment sorry 305 so
let's see indeed 305 so there you go so
they match up so again the rule is to
summarize your ke categories to K
dummies set up K minus 1 dummies and
include K minus 1 dummies along the
intercept or include all K dummies but
exclude the intercept short and sweet