1
00:00:11,080 --> 00:00:16,840
In this video, we will be introduced to the next section of this course, this section is about Victor

2
00:00:16,840 --> 00:00:20,180
auto regression, which is basically the victor version of Arima.

3
00:00:21,070 --> 00:00:22,760
So why do we want to do this?

4
00:00:23,590 --> 00:00:29,190
Well, in the previous sections, consider that we only ever looked at a single time series at a time.

5
00:00:29,800 --> 00:00:35,650
Our main assumption when doing so is that the future of the Time series is predictable, using only

6
00:00:35,650 --> 00:00:37,510
past values of that Time series.

7
00:00:37,990 --> 00:00:41,050
Now, one has to wonder, is this a valid assumption?

8
00:00:41,800 --> 00:00:46,380
Well, as with most real world problems, the answer is it depends.

9
00:00:46,990 --> 00:00:52,000
But you can see how in the real world one variable might have some effect on another.

10
00:00:52,540 --> 00:00:57,150
For example, suppose that the number of visitors to a website has increased.

11
00:00:57,670 --> 00:01:01,770
This might have an effect on the number of Web servers used to serve that website.

12
00:01:02,350 --> 00:01:08,260
Your auto scalar will create more machines in order to handle more traffic and both number of visitors

13
00:01:08,260 --> 00:01:10,120
and number of Web servers over time.

14
00:01:10,300 --> 00:01:11,470
Our Time series.

15
00:01:13,400 --> 00:01:19,340
Another example which is quite controversial is that the minimum wage has a negative effect on the number

16
00:01:19,340 --> 00:01:25,190
of jobs, the minimum wage at each point in time is a time series, as are employment levels.

17
00:01:25,670 --> 00:01:31,340
So given these real world situations, we know intuitively that at times there is maybe better predicted

18
00:01:31,550 --> 00:01:34,670
by other variables other than its own past values.

19
00:01:35,390 --> 00:01:39,860
In fact, it should make you question why did we use a rhema in the first place?

20
00:01:44,510 --> 00:01:49,710
In this section, we'll start by following the same basic outline of progressing from theory to code.

21
00:01:50,360 --> 00:01:55,070
You'll see that the theory for vector models is pretty straightforward and is basically just a vector

22
00:01:55,070 --> 00:01:57,080
version of what we saw before.

23
00:01:57,770 --> 00:02:02,840
Well, then look at how to use these vector models in code, which is similar to Arima, but perhaps

24
00:02:02,840 --> 00:02:06,460
a bit inconsistent after doing this.

25
00:02:06,620 --> 00:02:08,460
Well, then go back to looking at theory.

26
00:02:09,020 --> 00:02:11,330
This part is optional, but I think it's interesting.

27
00:02:12,080 --> 00:02:16,990
You'll see how amah models are really just special cases of the vector models we study here.

28
00:02:17,840 --> 00:02:20,510
We'll learn how to convert models from one form to another.

29
00:02:20,720 --> 00:02:25,640
And this should give you a deeper appreciation of what we've studied and how everything is related.