1
00:00:01,340 --> 00:00:04,830
So we Dandi logistic model with one predictive variable.

2
00:00:05,700 --> 00:00:06,840
And we got this result.

3
00:00:07,950 --> 00:00:10,380
We have the B does it all and B, the one values here.

4
00:00:11,130 --> 00:00:12,630
B, does it with zero point six one.

5
00:00:12,840 --> 00:00:15,420
And B, the one is minus zero point zero three.

6
00:00:17,250 --> 00:00:22,650
Once we get the values or B, does it want to be done, we can calculate the value of probability of

7
00:00:22,650 --> 00:00:24,570
Y is equal to one using the formula.

8
00:00:24,630 --> 00:00:26,820
I showed you in the last lecture.

9
00:00:30,120 --> 00:00:36,210
Now, when we ran the model in the output, we also get these other values, which is standard error,

10
00:00:36,570 --> 00:00:40,800
the value and the probability value, which is also called p value.

11
00:00:42,060 --> 00:00:43,740
I will talk about this briefly here.

12
00:00:47,260 --> 00:00:54,700
All these three values are calculated to establish whether this particular variable, that is the price

13
00:00:54,700 --> 00:00:59,620
variable is actually impacting the very well or not.

14
00:01:02,410 --> 00:01:04,050
Take a moment here and think about it.

15
00:01:05,020 --> 00:01:10,300
This variable will impact the response variable.

16
00:01:11,230 --> 00:01:14,620
Only if the beta corresponding to it is non-zero.

17
00:01:15,910 --> 00:01:20,200
If beta one is zero, whatever change we may make to this variable.

18
00:01:20,620 --> 00:01:22,270
Nothing will happen to the response variable.

19
00:01:24,310 --> 00:01:27,780
So we want to establish that beta one is non-zero.

20
00:01:29,320 --> 00:01:32,980
Mathematically began guided in this hypothesis form.

21
00:01:33,400 --> 00:01:40,300
That null hypothesis is there is no relationship between price and sored variable.

22
00:01:41,670 --> 00:01:44,080
The ordinate is that there is some relationship.

23
00:01:45,340 --> 00:01:46,750
We want to negate that.

24
00:01:46,750 --> 00:01:47,040
We do.

25
00:01:47,080 --> 00:01:54,190
One is zero so that we can establish that price is impacting the response variable.

26
00:01:57,220 --> 00:02:01,780
You may be thinking that we already got a non-zero value for this be done.

27
00:02:03,280 --> 00:02:04,720
So why are we discussing this?

28
00:02:05,320 --> 00:02:14,290
The point is that although we got an expected value of beta using the maximum likelihood method, how

29
00:02:14,290 --> 00:02:17,020
confident are we with this media value?

30
00:02:19,360 --> 00:02:26,200
There is still a chance that it was actually zero, but when we calculated using the maximum likelihood

31
00:02:26,200 --> 00:02:26,560
method.

32
00:02:27,880 --> 00:02:31,510
There is some probability that we are getting this value instead of zero.

33
00:02:33,250 --> 00:02:39,910
So these three values is helping us get that confidence level with which we can say that we do.

34
00:02:39,930 --> 00:02:41,140
One is non-zero.

35
00:02:43,270 --> 00:02:45,250
So the first column is standard error.

36
00:02:46,150 --> 00:02:52,510
Standard error is like standard deviation and represents the expected deviation of the actual return

37
00:02:53,230 --> 00:03:01,180
from these predicted ones to the predicted values zero point zero three and its deviation from the true

38
00:03:01,180 --> 00:03:08,350
value is nearly zero point zero one, corresponding to the standard error because it does devalue.

39
00:03:09,190 --> 00:03:15,130
Zewail You can be noted simply by dividing this estimated beta value by the standard error.

40
00:03:15,430 --> 00:03:22,180
So zero point zero three zero zero zero point zero one is nearly this two point forward value corresponding

41
00:03:22,180 --> 00:03:30,820
to this devalue, which is also called a Zis statistic using a Z statistic graph corresponding to the

42
00:03:30,820 --> 00:03:31,570
Z value.

43
00:03:31,720 --> 00:03:33,640
We can get a P value.

44
00:03:34,420 --> 00:03:40,660
P value is telling us the probability of BITA actually being zero when we are getting the predicted

45
00:03:40,660 --> 00:03:43,240
beta value as minus zero point zero three.

46
00:03:43,720 --> 00:03:51,340
And when the standard edit is zero point zero one, you may be Orvil in these last two lines.

47
00:03:54,990 --> 00:03:55,540
Don't worry.

48
00:03:55,820 --> 00:03:58,200
Here's the key takeaway from these values.

49
00:04:01,640 --> 00:04:06,810
Whenever we get all these values, we just need to check this P-value.

50
00:04:08,690 --> 00:04:12,020
This P-value should be lower than a threshold value.

51
00:04:13,640 --> 00:04:16,280
You can select this threshold as body or convenience.

52
00:04:17,780 --> 00:04:22,730
That threshold will give you the confidence level with which you can say that bidets actually impacting

53
00:04:22,730 --> 00:04:23,300
the response.

54
00:04:23,320 --> 00:04:23,970
Woodinville or not.

55
00:04:25,070 --> 00:04:29,960
Usually we have a threshold value of five percent or one percent.

56
00:04:31,160 --> 00:04:35,420
So as you can see that here, the P value is zero point zero zero six.

57
00:04:36,320 --> 00:04:44,480
This basically means that there is only a zero point zero six percent chance that the actual beta was

58
00:04:44,480 --> 00:04:48,410
zero and we predicted it to be zero point zero three.

59
00:04:50,600 --> 00:04:59,330
Therefore, just look at the P value of all the variables, whichever variable has a P value lower than

60
00:04:59,330 --> 00:05:00,320
a threshold value.

61
00:05:01,070 --> 00:05:05,600
You can state that all those variables are impacting the response variable.

62
00:05:07,280 --> 00:05:09,560
But others, you may not be sure.
