1
00:00:01,130 --> 00:00:05,735
Welcome back to Practical Time Series Analysis,

2
00:00:05,735 --> 00:00:07,755
and welcome to week five.

3
00:00:07,755 --> 00:00:14,580
We've been engaging the modeling process for an autoregressive or moving-average model.

4
00:00:14,580 --> 00:00:18,200
We've been trying to determine the appropriate order of the model,

5
00:00:18,200 --> 00:00:21,330
together with how to really estimate the coefficients

6
00:00:21,330 --> 00:00:26,150
that have your model describe the data as well as we can.

7
00:00:26,150 --> 00:00:30,330
What we're looking for this week is a measure of the quality, overall quality,

8
00:00:30,330 --> 00:00:35,395
of the model, and we'll use the Akaike information criterion.

9
00:00:35,395 --> 00:00:37,970
We'll look at some others as well.

10
00:00:37,970 --> 00:00:41,655
The idea here is similar to what you might have seen in regression.

11
00:00:41,655 --> 00:00:44,575
If you remember, in multivariate regression

12
00:00:44,575 --> 00:00:48,225
you will have an adjusted r squared term very often,

13
00:00:48,225 --> 00:00:51,620
that makes variables as they enter the model,

14
00:00:51,620 --> 00:00:53,835
pay a little bit of a tax.

15
00:00:53,835 --> 00:00:59,440
The idea is that we want to develop as simple as parsimonious model as possible.

16
00:00:59,440 --> 00:01:05,235
So, we explore these ideas through the Akaike Information criterion.

17
00:01:05,235 --> 00:01:07,320
We also try to increase the complexity of

18
00:01:07,320 --> 00:01:10,540
our modeling process by looking at mixed models.

19
00:01:10,540 --> 00:01:12,510
The ARMA models obviously,

20
00:01:12,510 --> 00:01:16,020
the autoregressive moving-average, together with ARIMA,

21
00:01:16,020 --> 00:01:19,420
integrated, autoregressive moving-average models,

22
00:01:19,420 --> 00:01:22,625
that allow us to deal with a trend in our data.

23
00:01:22,625 --> 00:01:28,455
We'll take at this point is a pretty powerful set of modeling ideas,

24
00:01:28,455 --> 00:01:33,680
and begin to apply them to some real-world datasets that we find interesting.

25
00:01:33,680 --> 00:01:35,000
Have a terrific week.