Interpreting y intercept in regression model

- [Instructor] Adrianna gathered data

on different schools' winning percentages

and the average yearly
salary of their head coaches

in millions of dollars in
the years 2000 to 2011.

She then created the following
scatter plot and trend line.

So this is salary in millions of dollars

and the winning percentage.

And so here, we have a coach
who made over $4 million,

and looks like they won
over 80% of their games.

But you have this coach over here

who has a salary of a little over

a million and a half dollars,

and they are winning over 85%,

and so each of one of these data points

is a coach,

and is plotting their salary

or their winning percentage
against their salary.

Assuming the line correctly
shows the trend in the data,

and it's a bit of an assumption,

there are some outliers here

that are well away from the model,

and this isn't a,

it looks like there's a linear,

a positive linear correlation here,

but it's not super tight

and there's a bunch of
coaches right over here,

in the lower salary area,

going all the way from
20 something percent

to over 60 percent.

Assuming the line correctly
shows the trend in the data,

what does it mean that the
line's y intercept is 39?

Well if you believe the model,

then the y intercept of being 39

would be the model is saying

that if someone makes no money,

that they could, zero dollars,

that they could win,

that the model would expect
them to win 39% of their games,

which seems a little unrealistic,

because you would expect most
coaches to get paid something.

But anyway, let's see
which of this choices

actually describe that.

So let me look at the choices.

The average salary was
39 million dollars, nope.

No one on our chart made 39 million.

On average, each million
dollar increase in salary

was associated with a 39%
increase in winning percentage.

That would be something
related to the slope

and the slope was definitely not 39.

The average winning percentage was 39%,

we know that wasn't the case either.

The model indicates
that teams with coaches

who had a salary of zero millions dollars

will average a winning
percentage of approximately 39%.

Yeah this is the closest statement

to what we just said,

that if you believe that model,

and that's a big if,

if you believe this model,

then this model says
someone making zero dollars

will get 39%,

and this is frankly why you
have to be skeptical of models.

They're not going to be perfect,

especially in extreme cases oftentimes,

but who knows.

Anyway, hopefully you found that useful.