WEBVTT

99:59:59.999 --> 99:59:59.999
♪ [music] ♪

99:59:59.999 --> 99:59:59.999
- [Thomas Stratmann] Hi!

99:59:59.999 --> 99:59:59.999
In the upcoming series of videos

99:59:59.999 --> 99:59:59.999
we're going to give you
a shiny new tool

99:59:59.999 --> 99:59:59.999
to put into your
Understanding Data toolbox:

99:59:59.999 --> 99:59:59.999
linear regression.

99:59:59.999 --> 99:59:59.999
Say you've got this theory.

99:59:59.999 --> 99:59:59.999
You've witnessed
how good-looking people

99:59:59.999 --> 99:59:59.999
seem to get special perks.

99:59:59.999 --> 99:59:59.999
You're wondering -- "Where else
might we see this phenomenon?"

99:59:59.999 --> 99:59:59.999
What about full professors?

99:59:59.999 --> 99:59:59.999
Is it possible
good-looking professors

99:59:59.999 --> 99:59:59.999
might get special perks too?

99:59:59.999 --> 99:59:59.999
Is it possible
students treat them better

99:59:59.999 --> 99:59:59.999
by showering them
with better student evaluations?

99:59:59.999 --> 99:59:59.999
If so, is the effect of looks

99:59:59.999 --> 99:59:59.999
on evaluation score
big or [inaudible]?

99:59:59.999 --> 99:59:59.999
And say there is a new professor
starting at the university.

99:59:59.999 --> 99:59:59.999
What can we predict
about his evaluation

99:59:59.999 --> 99:59:59.999
simply by his looks?

99:59:59.999 --> 99:59:59.999
Given that these evaluations
can determine pay raises,

99:59:59.999 --> 99:59:59.999
if this theory were true
we might see professors resort

99:59:59.999 --> 99:59:59.999
to some surprising tactics
to boost their scores.

99:59:59.999 --> 99:59:59.999
Suppose you wanted to find out

99:59:59.999 --> 99:59:59.999
if evaluations really improve
with better looks.

99:59:59.999 --> 99:59:59.999
How would you go about
testing this hypothesis?

99:59:59.999 --> 99:59:59.999
You could collect data.

99:59:59.999 --> 99:59:59.999
First you would have students rate
on a scale from 1 to 10

99:59:59.999 --> 99:59:59.999
how good looking a professor was,

99:59:59.999 --> 99:59:59.999
which gives you
an average beauty score.

99:59:59.999 --> 99:59:59.999
Then you could retrieve
the teacher's teaching evaluations

99:59:59.999 --> 99:59:59.999
from 25 students.

99:59:59.999 --> 99:59:59.999
Let's look at these two variables
at the same time

99:59:59.999 --> 99:59:59.999
by using a scatterplot.

99:59:59.999 --> 99:59:59.999
We'll put beauty
on the horizontal axis,

99:59:59.999 --> 99:59:59.999
and teacher evaluations
on the vertical axis.

99:59:59.999 --> 99:59:59.999
For example, this dot
represents Professor Peate,

99:59:59.999 --> 99:59:59.999
who received a beauty score of 3

99:59:59.999 --> 99:59:59.999
and an evaluation of 8.425.

99:59:59.999 --> 99:59:59.999
This one way out here
is Professor Helmchen.

99:59:59.999 --> 99:59:59.999
- [Professor Helmchen]
Ridiculously good-looking!

99:59:59.999 --> 99:59:59.999
- [Thomas] Who got
a very high beauty score,

99:59:59.999 --> 99:59:59.999
but not such a good evaluation.

99:59:59.999 --> 99:59:59.999
Can you see a trend?

99:59:59.999 --> 99:59:59.999
As we move from left to right
on the horizontal axis,

99:59:59.999 --> 99:59:59.999
from the ugly to the gorgeous,

99:59:59.999 --> 99:59:59.999
we see a trend upwards
in evaluation scores.

99:59:59.999 --> 99:59:59.999
By the way, the data
we're exploring in this series

99:59:59.999 --> 99:59:59.999
is not made up --
it comes from a real study

99:59:59.999 --> 99:59:59.999
done at the University of Texas.

99:59:59.999 --> 99:59:59.999
If you're wondering, "pulchritude"
is just the fancy academic way

99:59:59.999 --> 99:59:59.999
of saying beauty.

99:59:59.999 --> 99:59:59.999
With scatterplots
it can sometimes be hard

99:59:59.999 --> 99:59:59.999
to make out the exact relationship
between two variables --

99:59:59.999 --> 99:59:59.999
especially when the variables
bounce around quite a bit

99:59:59.999 --> 99:59:59.999
as we go from left to right.

99:59:59.999 --> 99:59:59.999
One way to cut through
this bounciness

99:59:59.999 --> 99:59:59.999
is to draw a straight line
through the data cloud

99:59:59.999 --> 99:59:59.999
in such a way that this line
summarizes the data

99:59:59.999 --> 99:59:59.999
as closely as possible.

99:59:59.999 --> 99:59:59.999
The technical term for this
is "linear regression."

99:59:59.999 --> 99:59:59.999
Later on we'll talk about
how this line is created,

99:59:59.999 --> 99:59:59.999
but for now we can assume
that the line fits the data

99:59:59.999 --> 99:59:59.999
as closely as possible.

99:59:59.999 --> 99:59:59.999
So, what can this line tell us?

99:59:59.999 --> 99:59:59.999
First, we immediately see

99:59:59.999 --> 99:59:59.999
if the line is sloping
upward or downward.

99:59:59.999 --> 99:59:59.999
In our data set we see
the [fitted] line slopes upward.

99:59:59.999 --> 99:59:59.999
It thus confirms what
we have conjectured earlier

99:59:59.999 --> 99:59:59.999
by just looking at the scatterplot.

99:59:59.999 --> 99:59:59.999
The upward slope means
that there is a positive association

99:59:59.999 --> 99:59:59.999
between looks
and evaluation scores.

99:59:59.999 --> 99:59:59.999
In other words, on average,

99:59:59.999 --> 99:59:59.999
better-looking professors
are getting better evaluations.

99:59:59.999 --> 99:59:59.999
For other data sets we might see
a stronger positive association.

99:59:59.999 --> 99:59:59.999
Or, you might see
a negative association.

99:59:59.999 --> 99:59:59.999
Or perhaps no association at all.

99:59:59.999 --> 99:59:59.999
And our lines
don't have to be straight.

99:59:59.999 --> 99:59:59.999
They can curve to fit the data
when necessary.

99:59:59.999 --> 99:59:59.999
This line also gives us
a way to predict outcomes.

99:59:59.999 --> 99:59:59.999
We can simply take a beauty score
and read off the line

99:59:59.999 --> 99:59:59.999
what the predicted
evaluation score would be.

99:59:59.999 --> 99:59:59.999
So, back to our new professor.

99:59:59.999 --> 99:59:59.999
- [Professor Lloyd] Look familiar?

99:59:59.999 --> 99:59:59.999
- [Thomas] We can precisely
predict his evaluation score.

99:59:59.999 --> 99:59:59.999
"But wait! Wait!" you might say.

99:59:59.999 --> 99:59:59.999
"Can we trust this prediction?"

99:59:59.999 --> 99:59:59.999
How well does
this one beauty variable

99:59:59.999 --> 99:59:59.999
really predict evaluations?

99:59:59.999 --> 99:59:59.999
Linear regression gives us
some useful measures

99:59:59.999 --> 99:59:59.999
to answer those questions

99:59:59.999 --> 99:59:59.999
which we'll cover
in a future video.

99:59:59.999 --> 99:59:59.999
We also have to be aware
of other pitfalls

99:59:59.999 --> 99:59:59.999
before we draw
any definite conclusions.

99:59:59.999 --> 99:59:59.999
You could imagine a scenario

99:59:59.999 --> 99:59:59.999
where what is driving
the association

99:59:59.999 --> 99:59:59.999
we see is really a third variable
that we have left out.

99:59:59.999 --> 99:59:59.999
For example,
the difficulty of the course

99:59:59.999 --> 99:59:59.999
might be behind
the positive association

99:59:59.999 --> 99:59:59.999
between beauty ratings
and evaluation scores.

99:59:59.999 --> 99:59:59.999
Easy intro. courses
get good evaluations.

99:59:59.999 --> 99:59:59.999
Harder, more advanced courses
get bad evaluations.

99:59:59.999 --> 99:59:59.999
And younger professors might
get assigned to intro. courses.

99:59:59.999 --> 99:59:59.999
Then, if students judge
younger professors more attractive,

99:59:59.999 --> 99:59:59.999
you will find
a positive association

99:59:59.999 --> 99:59:59.999
between beauty ratings
and evaluation scores.

99:59:59.999 --> 99:59:59.999
But it's really
the difficulty of the course.

99:59:59.999 --> 99:59:59.999
The variable that we've left out,
not beauty,

99:59:59.999 --> 99:59:59.999
that is driving evaluation scores.

99:59:59.999 --> 99:59:59.999
In that case, all the primping
would be for naught --

99:59:59.999 --> 99:59:59.999
a case of mistaken correlation
for causation,

99:59:59.999 --> 99:59:59.999
something we'll talk about further
in a later video.

99:59:59.999 --> 99:59:59.999
And what if there were
other important variables

99:59:59.999 --> 99:59:59.999
that affect both beauty ratings
and evaluation scores?

99:59:59.999 --> 99:59:59.999
You might want to add
considerations like skill,

99:59:59.999 --> 99:59:59.999
race, sex, and whether English
is the teacher's native language

99:59:59.999 --> 99:59:59.999
to isolate more cleanly the effect
of beauty on evaluations.

99:59:59.999 --> 99:59:59.999
When we get
into multiple regression

99:59:59.999 --> 99:59:59.999
we will be able to measure
the impact of beauty

99:59:59.999 --> 99:59:59.999
on teacher evaluations

99:59:59.999 --> 99:59:59.999
while accounting
for other variables

99:59:59.999 --> 99:59:59.999
that might confound
this association.

99:59:59.999 --> 99:59:59.999
Next up, we'll get our hands dirty
by playing with this data

99:59:59.999 --> 99:59:59.999
to gain a better understanding
of what this line can tell us.

99:59:59.999 --> 99:59:59.999
- [Narrator] Congratulations!

99:59:59.999 --> 99:59:59.999
You're one step closer
to being a data ninja!

99:59:59.999 --> 99:59:59.999
However, to master this you'll need
to strengthen your skills

99:59:59.999 --> 99:59:59.999
with some practice questions.

99:59:59.999 --> 99:59:59.999
Ready for your next mission?
Click "Next Video."

99:59:59.999 --> 99:59:59.999
Still here?

99:59:59.999 --> 99:59:59.999
Move from understanding eata
to understanding your world

99:59:59.999 --> 99:59:59.999
by checking out MRU's
other popular economics videos.

99:59:59.999 --> 99:59:59.999
♪ [music] ♪