1
00:00:00,920 --> 00:00:07,230
In this video, we will discuss one of the problems we may face, if not more than this problem is called

2
00:00:07,410 --> 00:00:08,730
Petros Clandestinity.

3
00:00:09,940 --> 00:00:14,560
If you remember in the lecture in which we found out the confidence interval of Bita.

4
00:00:15,900 --> 00:00:24,030
I told you that this dumb sigma squared, which is the variance in the error term, is helping us find

5
00:00:24,030 --> 00:00:26,100
out the confidence interval of beta.

6
00:00:27,930 --> 00:00:33,840
There was one underlying assumption regarding the Sigma squared, based on which we calculated the standard

7
00:00:33,840 --> 00:00:34,100
errors.

8
00:00:34,170 --> 00:00:35,730
Confidence intervals, etc..

9
00:00:36,980 --> 00:00:40,730
The assumption is that this variance value, it's constant.

10
00:00:42,730 --> 00:00:49,150
Unfortunately, it is often the case that variances of the error terms are non constant.

11
00:00:50,530 --> 00:00:55,890
For instance, the variances of error terms may increase with the value of the response that is deemed

12
00:00:55,900 --> 00:00:59,120
by variable GraphicLy.

13
00:00:59,500 --> 00:01:05,080
We can plot Deidre's do against Y to identify the non constant variance of dieters.

14
00:01:07,290 --> 00:01:13,650
As you can see in this graph, this funnel like shape is suggesting non constant variance.

15
00:01:15,380 --> 00:01:18,170
This issue is known as a cadastre city.

16
00:01:19,190 --> 00:01:24,380
You can see that the magnitude of residuals tends to increase with defeated values.

17
00:01:24,590 --> 00:01:31,870
Why so variance at this point is limited within this range.

18
00:01:33,250 --> 00:01:38,430
Whereas variants here is larger as the boundaries are for the really.

19
00:01:40,210 --> 00:01:42,490
Now, how to handle heteros, cadastre city.

20
00:01:43,920 --> 00:01:49,110
When you faced this problem, where did variances increasing with larger footed values of why?

21
00:01:50,150 --> 00:01:52,370
We need to scale down these large values of a.

22
00:01:53,530 --> 00:02:01,170
So we can use functions like logo fly or under route way so that larger, well lives away are shut down.

23
00:02:02,450 --> 00:02:05,540
Thus leading to reduction in heteros congested city.

24
00:02:07,670 --> 00:02:10,280
You can see the graph of residuals against Longway.

25
00:02:11,830 --> 00:02:15,050
Do they do it now appear to have constant variants?

26
00:02:16,470 --> 00:02:20,640
Although there is some evidence of a slight non-linear relationship in the data.

27
00:02:23,200 --> 00:02:28,870
So this is how we identify, if that is heteros contested city in the model, and if there is, this

28
00:02:28,870 --> 00:02:29,740
is how we handle it.
