M17 Power Analysis 1 Proportion Video Lecture | Mintab Tutorials for Beginners - Data & Analytics

being the brilliant surgeon that I am I
have actually developed a new technique
for performing splenectomy x' I know
it's amazing I'm incredible it seems to
be a great way to do it except except I
have critics who say that my method is
causing a greater mortality rate or a
greater percentage of people are dying
on the table because of my technique
than those who have the traditional
splenectomy so we want to find out if
this is true so a study has been
proposed using my technique and then
we're going to and we're going to
compare it to what is the known
mortality rate for the traditional
technique now the known mortality rate
is up here you'll see the average
mortality rate for splenectomy x'
nationwide is 0.02 6 so zero point zero
to six is the is the mortality rate
which means the prop the proportion of
people who died on the table
during the traditional splenectomy
surgery we're gonna find out or we're
gonna try and design a study to find out
if there's a greater mortality rate in
mine we know some things about our study
okay so how large a sample would be
needed to determine with the
significance of 0.05 so we want to work
at 95% confidence or significance of
0.05
if the mortality rate of my surgery were
greater than zero point zero to six we
want to have a 90% chance so this is
called our power to notice a difference
between point zero three zero and point
zero two six
let me explain power again a little bit
remember power is your probability of
detecting or of avoiding a type two
error if there really is a difference
between the two
what is your chance of detecting it well
it depends on how far apart the
difference is so you have to say what
difference you're talking about and here
I'm talking about a difference between
point zero three zero and point zero two
six and I want to have a ninety percent
chance of detecting a difference
assuming it's that large or larger
okay so let's go to how we do this which
is again in our stat menu but this time
we're doing what's called a power
analysis my power is 90% or 0.9 so when
I do this if you look at the list you
could probably see where we're gonna go
we're gonna go to power and sample size
we're actually doing a power analysis
now I have to know what kind of
hypothesis test this would be so I know
which kind of power analysis to do this
would be a one proportion test I mean
have one sample and I'm testing the
proportion of people who died on the
table
mortality rate so I'm going to go to one
proportion and now I need to fill in two
of these and it's gonna tell me the
other one well I don't know sample size
that's what I'm actually trying to find
out how many people would I need to have
in this study so that I could do detect
this difference with a 90% chance so I'm
going to leave sample size blank that's
what I want to know comparison
proportion so the comparison proportion
is the proportion that we're trying to
look for that could be knew the zero
point zero three zero zero point zero
three zero the power value I wanted to
have a power of 90 percent 0.9 0.9 and
the hypothesized proportion is what the
old one was what the control was what
the traditional one was zero point zero
two six zero point zero two six now
we're going to click on options before
we go on and we're going to just check
these things the alternative hypothesis
in this case is actually a greater than
so I need to switch that the
significance is fine so I'm going to say
okay and say okay and Minitab performs a
power analysis for us and you're gonna
see that if I actually wanted to be able
to detect a difference of that small
between point zero three zero and point
zero two six I would need and have
ninety percent chance of doing it I
would need fourteen thousand four
hundred and twenty three people to
undergo the surgery based on my new
technique before we could make a
determination that
fine that's probably unrealistic you're
not going to let 14,000 people have a
new surgery that might be causing more
deaths you're going to probably going to
either have to accept a lower power
value or a larger difference between the
two so that you can bring that number
down before you can start to make a
determination