1
00:00:04,500 --> 00:00:07,830
Let us comment on the
results of the study.

2
00:00:07,830 --> 00:00:13,020
Robert Parker predicted that the
1986 Bordeaux wine is very good

3
00:00:13,020 --> 00:00:16,030
to sometimes exceptional.

4
00:00:16,030 --> 00:00:19,220
On the other hand,
Ashenfelter said

5
00:00:19,220 --> 00:00:22,240
that the 1986 Bordeaux
wine is mediocre.

6
00:00:22,240 --> 00:00:25,340
And made the prediction
that the 1989 Bordeaux would

7
00:00:25,340 --> 00:00:28,570
be the wine of the century
and the 1990 Bordeaux

8
00:00:28,570 --> 00:00:30,570
would be even better.

9
00:00:30,570 --> 00:00:34,340
In wine options, the
1989 Bordeaux wine

10
00:00:34,340 --> 00:00:38,170
sold for more than
twice the price of 1986.

11
00:00:38,170 --> 00:00:42,820
And the 1990 Bordeaux wine
sold for even higher prices.

12
00:00:42,820 --> 00:00:46,690
Later, Ashenfelter predicted
that the 2000 and the 2003

13
00:00:46,690 --> 00:00:49,130
Bordeaux wines would be great.

14
00:00:49,130 --> 00:00:52,340
And in this case,
Robert Parker stated

15
00:00:52,340 --> 00:00:56,580
the 2000 is the greatest vintage
Bordeaux has ever produced,

16
00:00:56,580 --> 00:00:57,830
in agreement with Ashenfelter.

17
00:01:03,200 --> 00:01:07,390
So what is the analytics
edge in this case?

18
00:01:07,390 --> 00:01:10,690
What we have developed is
a linear regression model,

19
00:01:10,690 --> 00:01:15,460
a simple but rather
powerful model

20
00:01:15,460 --> 00:01:17,350
for predicting quality of wines.

21
00:01:17,350 --> 00:01:20,180
It only used few
variables and we

22
00:01:20,180 --> 00:01:23,789
have seen that it predicted
wine prices quite well.

23
00:01:23,789 --> 00:01:27,110
In fact, in many
cases it outperformed

24
00:01:27,110 --> 00:01:30,060
wine expert's opinions.

25
00:01:30,060 --> 00:01:33,710
And what is impressive, in
this first introductory lecture

26
00:01:33,710 --> 00:01:38,490
to linear regression, is that
an analytics approach that

27
00:01:38,490 --> 00:01:43,800
uses data to build a model
that improves decision making

28
00:01:43,800 --> 00:01:46,960
is effective in a traditionally
qualitative problem.