1
00:00:01,990 --> 00:00:05,276
What is the relationship between
Attack and Health Points?

2
00:00:07,981 --> 00:00:12,442
There are two types of plots for
charting out relationships in Seaborn.

3
00:00:12,442 --> 00:00:15,302
Scatterplots and lineplots.

4
00:00:15,302 --> 00:00:16,964
Let's start with a scatterplot.

5
00:00:19,613 --> 00:00:22,628
Seaborn uses a declarative
approach to plotting,

6
00:00:22,628 --> 00:00:26,282
which means that the function
calls are self-descriptive.

7
00:00:26,282 --> 00:00:29,343
To make a scatterplot,

8
00:00:29,343 --> 00:00:33,934
we call the scatterplot function,

9
00:00:33,934 --> 00:00:40,820
sns.scatterplot, and
it takes three parameters,

10
00:00:40,820 --> 00:00:46,031
data=pokemon, x=HP and y=Attack.

11
00:00:46,031 --> 00:00:49,103
Shift+Enter to run the cell.

12
00:00:49,103 --> 00:00:53,604
The scatterplot shows us that
there may be a general positive

13
00:00:53,604 --> 00:00:57,763
correlation between HP and
Attack, with one outlier.

14
00:00:57,763 --> 00:01:01,961
Generally as HP increases, so does Attack.

15
00:01:01,961 --> 00:01:06,851
Pokemon with larger health
points tend to be stronger.

16
00:01:06,851 --> 00:01:10,310
Now let's try a lineplot.

17
00:01:10,310 --> 00:01:14,791
We call the lineplot function to make one.

18
00:01:14,791 --> 00:01:17,706
sns.lineplot.

19
00:01:17,706 --> 00:01:23,646
It takes the same three parameters,

20
00:01:23,646 --> 00:01:30,553
data=Pokemon, x=HP, and y= Attack.

21
00:01:30,553 --> 00:01:33,841
Shift+Enter to run the cell.

22
00:01:33,841 --> 00:01:38,635
The lineplot doesn't do a great job
at showing us information that can be

23
00:01:38,635 --> 00:01:39,901
easily inferred.

24
00:01:39,901 --> 00:01:44,326
Recall, that a lineplot is
better at showing an x-axis that

25
00:01:44,326 --> 00:01:47,461
follows a continuous variable like time.

26
00:01:47,461 --> 00:01:51,200
In this example,
we're plotting out a discrete variable HP.

27
00:01:52,770 --> 00:01:56,544
So what happens is that the lineplot
goes all over the place.

28
00:01:56,544 --> 00:02:02,422
And it's harder to infer a trend compared
to the scatterplot, that's okay.

29
00:02:02,422 --> 00:02:07,320
A part of exploratory data analysis
is trying out different things to see

30
00:02:07,320 --> 00:02:08,531
what works well.

31
00:02:08,531 --> 00:02:14,122
It's normal for some plots to show
us better insights than other kinds.

32
00:02:14,122 --> 00:02:17,082
So let's explore more of the scatterplot.

33
00:02:17,082 --> 00:02:22,027
We can introduce the hue keyword to look
at a breakdown of how different types

34
00:02:22,027 --> 00:02:24,512
are distributed in the scatterplot.

35
00:02:24,512 --> 00:02:30,259
Remember that the type of Pokemon is
a categorical variable like water,

36
00:02:30,259 --> 00:02:31,962
grass and electric.

37
00:02:31,962 --> 00:02:38,139
sns.scatterplot, data=pokemon,

38
00:02:38,139 --> 00:02:41,511
x = HP, y = Attack.

39
00:02:47,883 --> 00:02:49,736
And hue = Type.

40
00:02:54,152 --> 00:02:56,797
I've got an AttributeError,

41
00:02:56,797 --> 00:03:02,296
module 'seaborn' has no attribute
'scatterplt' with no o.

42
00:03:02,296 --> 00:03:06,569
So let's fix that typo scatterplot, and

43
00:03:06,569 --> 00:03:11,342
now I've got a scatterplot
with color codes.

44
00:03:11,342 --> 00:03:17,462
And this takes us into analyzing
the categorical aspects of our data.

45
00:03:17,462 --> 00:03:22,226
We can break this scatterplot down
further by using the relplot function and

46
00:03:22,226 --> 00:03:25,279
introducing the "col"
Column keyword argument,

47
00:03:25,279 --> 00:03:28,789
to make many scatterplots
based on the type of Pokemon.

48
00:03:32,032 --> 00:03:37,284
Sns.relplot is short for
relationship plot.

49
00:03:37,284 --> 00:03:41,664
Starts with the same parameters

50
00:03:41,664 --> 00:03:47,185
data=pokemon, x=HP and y= Attack.

51
00:03:50,542 --> 00:03:55,593
We will pass it the hue
keyword argument for Type, and

52
00:03:55,593 --> 00:04:00,990
now the "col" column keyword
argument which also takes Type

53
00:04:04,124 --> 00:04:07,400
Shift+Enter to run the cell.

54
00:04:07,400 --> 00:04:11,603
And right now it's really
difficult to see these new plots.

55
00:04:11,603 --> 00:04:15,944
So we'll add one more parameter,

56
00:04:15,944 --> 00:04:22,540
the col_wrap keyword argument, col_wrap=3.

57
00:04:26,513 --> 00:04:28,871
And now we have a scatterplot for

58
00:04:28,871 --> 00:04:34,560
each type of Pokemon that describes
the relationship between HP and Attack.

59
00:04:34,560 --> 00:04:39,181
It's nice to have a breakdown of types
like this because for some types,

60
00:04:39,181 --> 00:04:41,392
there are very few observations.

61
00:04:41,392 --> 00:04:46,084
Looking at the types that have many
observations, we can infer that generally,

62
00:04:46,084 --> 00:04:49,120
HP and
Attack are somewhat positively correlated.

63
00:04:50,460 --> 00:04:53,573
Now try practicing with
the defense style of Pokemon.

64
00:04:53,573 --> 00:04:57,740
Can you infer a relationship
between HP and defense?

65
00:05:02,448 --> 00:05:04,843
Before we move on to making more plots,

66
00:05:04,843 --> 00:05:08,047
let's write down our
observations to record them.

67
00:05:10,441 --> 00:05:16,090
We'll use a mark down cell and
say that generally, HP and

68
00:05:16,090 --> 00:05:23,045
Attack are somewhat correlated,
somewhat positively correlated.

69
00:05:26,631 --> 00:05:33,260
Pokemon with more HP tend to be stronger.

70
00:05:35,543 --> 00:05:36,160
Awesome.