Working with StatCrunch: Means Plots (With Confidence Intervals) Video Lecture | Statcrunch Tutorials for Beginners - Data & Analytics

Hi, my name's Bart Poulson and in this tutorial,
I'm going to show how to use StatCrunch, an
online statistics program, to calculate what
is called a mean's plot or means with confidence
intervals, uh, for quantitative variable broken
down by a categorical variable. The first
thing you need to do is get to statcrunch.com.
The second thing you need to do is to log
into your account. And now, we need to go
get the data set that we're going to use.
I'm gonna be using one that's already on StatCrunch,
so I come up here to where it says "Explore"
and I click on the little triangle here, and
I open it up where it says "Data." Now, I'm
going to use a data set that I've used before,
so it's going to pop up as soon as I type
it. I'm gonna use one called "yms Table 1.15."
I hit "Search," and it brings up this window.
It looks like a high school right here. I
click on that, and I have a data set with
51 rows, one row for each state and the District
of Columbia and several variables here. Um,
one that I'm going to look at is SAT math
scores by region. Now, to get a means plot,
what I do is I come over here to "Graphics"
and I go down to "Means Plot."
I select the outcome variable that I'm interested
in, which in this case is SAT math scores.
Obviously, I could do more than one if I just
held down the control-key or the apple-key
on my Mac and selected each of these. But
I'm just going to do one now to keep it a
little simple. Also, because if I do more
than one, it makes it harder to, uh, put titles
and labels on the charts. And I'm going to
group it by region. I hit the next one, and
I'm going to put some labels on it. The title
should say [typing] "Average SAT Math Scores
by Region of US." Ok. The x-axis, and we'll
put as [typing] "Average SAT—" oops
"—Math Score." And the, um, y-axis
will be "Region of US." Ok, I'm gonna ignore
all of these because I'm not doing, uh, many
modifications, and I don't need this because
I'm only doing one graph. So I press "Create
Graph." And tada. There we go. I'll make this
a little bigger for a second. Alright. What
we have here is a quantitative variable across
the bottom. It's the average SAT math score
for each state or locality in the data set.
These are the regions it is broken down, and
I have to admit, I don't know what they are,
but it would be a pretty simple thing to go
back and compare them with the names of the
states, um, in each one of 'em. I know that
this one, MTN, is mountain, and it includes
Utah, Colorado, and a few others. And what
you have for each one of these, uh, is a dot
that represents the mean for that particular
group. So for instance, the WSC right—
well, actually, let me use mountain. So for
mountain, the mean on the SAT Math is right
about 540. It's in the middle of the pack.
The highest mean is for WNC, the lowest mean
is for SA. That might be South Atlantic. Uh,
others are in between. Now, you can compare
the means, that's pretty easy. But the other
thing you wanna look at here are the confidence
intervals. That's these lines that go up and
down. Um, and what a confidence interval indicates,
is the probable range, uh, for the population
mean. The idea here is that we have data from
a particular sample, but, you know, there's
variability within the sample, and if we were
to sample again from the same population—and
that could be a little tricky to specify exactly
what the same population is—but if
we were to sample from the same population,
we might get means as high as this, or as
low as this. And this gives you a good feel
for the variability, um, in the data, and
also a good way to classify groups. So, for
instance, these ones right down here are obviously
lower than others because their confidence
intervals don't even come close to crossing
over with these ones. Uh, this one right here
has a very broad confidence interval and so
even though it's higher than this one right
here, the next one up, because there's so
much overlap, they're functionally equivalent.
It could just be a little fluke. The same
thing with this one, and with this one. Anyhow,
confidence intervals is an excellent way to
look at these scores on a quantitative variable
and to make an eyeball version of an inferential
test, uh, something that we may cover more
at a later date.