1
00:00:01,400 --> 00:00:07,670
So you would have noticed that there is one value of this outpoured that we have still not discussed

2
00:00:07,670 --> 00:00:13,490
and that value is called F statistic, and in this video we'll be discussing about F statistic.

3
00:00:15,580 --> 00:00:19,300
Now, if you see the result for each of the coefficient.

4
00:00:20,240 --> 00:00:26,630
There is a corresponding P-value, which is telling us whether each individual predictor is related

5
00:00:26,630 --> 00:00:31,040
to the response, I'm talking about these P values and probability values.

6
00:00:31,070 --> 00:00:31,910
These two columns.

7
00:00:34,500 --> 00:00:42,780
So since at least one of them has a small p value here, we have six of them to, since at least one

8
00:00:42,780 --> 00:00:50,340
of them has a small P value corresponding to their T value, we may think that shortly at least one

9
00:00:50,340 --> 00:00:52,830
of the predictors are related to the response.

10
00:00:54,810 --> 00:01:00,930
Therefore, we will say that there is a relationship between this point and political variables that

11
00:01:00,930 --> 00:01:06,660
is not all of the conditions of the variables, that is the we do have to be to be.

12
00:01:07,890 --> 00:01:10,520
Ardito, so all of them are not Dettol.

13
00:01:12,910 --> 00:01:19,960
So this argument may sound okay, but it is actually not, especially if we have a large number of variables.

14
00:01:21,830 --> 00:01:30,200
Let me put it mathematically for you first, what I'm saying is the null hypothesis is that all of them

15
00:01:30,200 --> 00:01:30,710
are zero.

16
00:01:32,180 --> 00:01:35,130
I want to prove that this is not the case.

17
00:01:35,510 --> 00:01:39,500
I want to say that at least one of the Beatles is definitely not.

18
00:01:40,430 --> 00:01:45,350
That is the predictor that I am selected, have some relationship with the response.

19
00:01:49,310 --> 00:01:58,050
To now, suppose if my model is saying that any one of the predictors say here we got sakes, suppose

20
00:01:58,050 --> 00:02:05,520
it was giving that only one of them to room number is significantly littered with Osprey's, which is

21
00:02:05,520 --> 00:02:07,890
the P value and the rest are not.

22
00:02:09,850 --> 00:02:19,540
Now, can I say with confidence that since bit of room number is significantly limited, I am confident

23
00:02:19,540 --> 00:02:23,950
that the predictors in this model are impacting the response variable.

24
00:02:25,430 --> 00:02:27,800
The answer is no, I cannot say that.

25
00:02:28,780 --> 00:02:37,780
Let me show you how to suppose I toss a coin five times and I say that, that if I get five heads in

26
00:02:37,780 --> 00:02:41,470
a row, I will turn that coin as a biased coin.

27
00:02:43,220 --> 00:02:50,810
Now, if this is my criteria of saying that the coin is biased, what is the probability that I will

28
00:02:51,080 --> 00:02:54,140
classify a fair coin as a biased coin?

29
00:02:56,420 --> 00:03:02,870
As you can see, probability of head in one toss of a coin is half.

30
00:03:04,920 --> 00:03:12,510
So what is the probability that in all the five doses, I'll get all the heads half into half into half

31
00:03:12,510 --> 00:03:19,380
and half and half five times, that is harvestable fail, which comes out three point zero three one

32
00:03:19,380 --> 00:03:19,860
two fail.

33
00:03:21,320 --> 00:03:30,380
This means that there is nearly three percent chance that I will wrongly classify a fair coin as a biased

34
00:03:30,380 --> 00:03:30,700
coin.

35
00:03:33,330 --> 00:03:37,770
Now, if I toss 100 fair coins five times.

36
00:03:39,530 --> 00:03:45,680
What is the probability of at least one of them getting all five head?

37
00:03:47,670 --> 00:03:53,730
So these are 100 coins, the probability of getting ahead is point five again.

38
00:03:55,430 --> 00:04:00,830
What is the probability that at least one of them is wrongly classified as a biased coin?

39
00:04:02,180 --> 00:04:08,690
So this we can calculate using this formula that I've shown here, it is first finding out the probability

40
00:04:08,690 --> 00:04:14,930
of getting all heads and none of these and we are finding one minus that probability.

41
00:04:15,380 --> 00:04:17,780
And that is coming out to be 95 percent.

42
00:04:18,730 --> 00:04:22,770
You do not need to know how this came out, but the result is important.

43
00:04:23,780 --> 00:04:30,530
The point that I am trying to make is, although the probability of calling a fair coin as a bad coin

44
00:04:30,890 --> 00:04:32,660
was only three percent.

45
00:04:33,750 --> 00:04:41,490
But if you are doing this experiment 100 times, it is almost certain that you will wrongly classify

46
00:04:41,670 --> 00:04:43,410
at least one of the Queen's.

47
00:04:45,440 --> 00:04:49,160
By almost certain, I mean, there is more than 95 percent probability.

48
00:04:50,640 --> 00:04:58,940
You know, this is similar to this statistic in the sense I said that we are 95 percent sure that we

49
00:04:58,970 --> 00:05:00,280
that there is not zero.

50
00:05:00,780 --> 00:05:03,120
When I saw it devalue and P-value.

51
00:05:04,910 --> 00:05:07,000
So we are nearly five percent unsure.

52
00:05:08,230 --> 00:05:14,260
So there is a five percent chance of a beta, which is zero being classified as a non-zero beta.

53
00:05:16,640 --> 00:05:18,760
Not if we have a large number of variables.

54
00:05:20,110 --> 00:05:26,890
All having five percent chance of getting wrongly guessed as significant, it is almost certain that

55
00:05:26,890 --> 00:05:31,690
at least one of them is actually wrongly saying that we have got that very well known.

56
00:05:33,640 --> 00:05:41,110
So if we depend on individual P values, there is a very high chance that we will incorrectly conclude

57
00:05:41,260 --> 00:05:45,100
that there is a relationship between predictor variables and irresponsible.

58
00:05:46,950 --> 00:05:52,380
The solution to this problem is to adjust for the number of predictors and then find the p value.

59
00:05:53,930 --> 00:05:58,700
This new statistic, which adjust just for the number of predictors, is called a statistic.

60
00:06:00,340 --> 00:06:04,600
This includes the number of variables in the form of this be.

61
00:06:06,570 --> 00:06:14,580
So all this then we run a multiple regression model, give a statistic value for the model and the corresponding

62
00:06:14,580 --> 00:06:15,180
P value.

63
00:06:16,770 --> 00:06:23,490
So this P-value must be checked against the threshold that whether there is a relationship between model

64
00:06:23,490 --> 00:06:30,840
predictors and the response, if there is, then we will look at the disvalue and P-value of each individual

65
00:06:30,840 --> 00:06:34,780
variable to identify which of the variables is significantly.

66
00:06:37,400 --> 00:06:38,820
I hope you understand the concept.

67
00:06:39,020 --> 00:06:40,730
I'll briefly summarize it.

68
00:06:42,110 --> 00:06:46,790
The idea is, instead of looking at individual variables.

69
00:06:47,960 --> 00:06:54,860
And saying that the model predictors are significantly impacting the response variable, we will look

70
00:06:54,860 --> 00:06:59,990
at a different statistic which takes into account the number of variables that we have.

71
00:07:01,830 --> 00:07:09,300
This ensures that we do not make the mistake of saying that the models have a significant relationship

72
00:07:09,300 --> 00:07:10,060
with predictor.

73
00:07:10,470 --> 00:07:16,830
If we got that, we are not to worry about positively impacting the response by chance.

74
00:07:17,910 --> 00:07:24,360
So to avoid that chant, we are using this new statistic, which is called a statistic, we look at

75
00:07:24,360 --> 00:07:28,110
its value and we look at the corresponding P value.

76
00:07:28,590 --> 00:07:33,890
If this P value is lower than the threshold value, say, one percent or five percent.

77
00:07:33,900 --> 00:07:40,040
Again, we say that the model predictors are significantly impacting the response.

78
00:07:40,500 --> 00:07:48,000
After that, we will look at the individual predictors and their estimate and P values, although not

79
00:07:48,330 --> 00:07:49,620
like I have told you earlier.

80
00:07:49,620 --> 00:07:56,370
Also, even if you do not understand the whole concept behind this statistic, the concept that you

81
00:07:56,370 --> 00:08:04,530
should take away from this video lesson is just that you have to check whether your P value of F statistic

82
00:08:04,530 --> 00:08:06,610
is lower than a threshold of five percent.

83
00:08:07,410 --> 00:08:11,700
You do not need to understand all of the concept behind Aphoristic.

84
00:08:12,810 --> 00:08:20,880
This will ensure that the model predicts that you used are having some significant impact on the response

85
00:08:20,880 --> 00:08:21,270
that responsivity will.