Lecture
According to the central limit theorem, the law of the distribution of a sum of a sufficiently large number of independent random variables (subject to certain non-rigid constraints) is arbitrarily close to normal.
Practically the central limit theorem can be used even when it comes to the sum of a relatively small number of random variables. When summing up independent random variables comparable in their dispersion, with an increase in the number of terms, the law of the distribution of the sum very soon becomes approximately normal. In practice, it is generally widely used approximate replacement of some laws of distribution of others; with the relatively low accuracy required from probabilistic calculations, such a replacement can also be made very approximately. Experience shows that when the number of terms is of the order of ten (and often less), the law of the distribution of the sum can usually be replaced by a normal one.
In practical problems, the central limit theorem is often used to calculate the probability that the sum of several random variables is within the specified limits.
Let be - independent random variables with mathematical expectations
and dispersions
.
Suppose that the conditions of the central limit theorem are satisfied (the values comparable in order of their influence on the dispersion of the sum) and the number of terms enough for the law of distribution of magnitude
(13.9.1)
could be considered approximately normal.
Then the probability that a random variable falls within the plot expressed by the formula
, (13.9.2)
Where , - expectation and standard deviation of magnitude - normal distribution function.
According to the theorems of addition of mathematical expectations and variances
(13.9.3)
Thus, in order to approximately find the probabilities of the sum of a large number of random variables falling on a given segment, it is not necessary to know the laws of the distribution of these quantities; it is enough to know only their characteristics. Of course, this applies only to the case when the main condition of the central limit theorem is satisfied - the uniformly small influence of the terms on the dispersion of the sum.
In addition to formulas of the type (13.9.2), in practice, formulas are often used in which instead of the sum of random variables their normalized figure appears
. (13.9.4)
Obviously
; .
If the distribution law close to normal with parameters (13.9.3), then the distribution law close to normal with parameters , . From here
. (13.9.5)
Note that the central limit theorem can be applied not only to continuous, but also to discrete random variables, provided that we will operate not with densities, but with distribution functions. Indeed, if the values discrete, their sum - also a discrete random variable and therefore, strictly speaking, cannot obey the normal law. However, all formulas of the type (13.9.2), (13.9.5) remain in force, since they do not include the density, but the distribution function. It can be proved that if discrete random variables satisfy the conditions of the central limit theorem, then the distribution function of their normalized sum (see formula (13.9.4)) when increasing unboundedly approaches the normal distribution function with parameters , .
A special case of the central limit theorem for discrete random variables is the Laplace theorem.
If produced independent experiences in each of which the event appears with probability then the ratio is fair
, (13.9.6)
Where - the number of occurrences at experiences, .
Evidence. Let produced independent experiments, in each of which with probability event may appear . Imagine a random variable. - the total number of occurrences of the event in experiences - as a sum
, (13.9.7)
Where - the number of occurrences at m experience.
According to proven in 13.8 of the theorem, the distribution law of the sum of identically distributed terms with an increase in their number approaches the normal law. Therefore, with a sufficiently large the formula (13.9.5) is valid, where
. (13.9.8)
AT 10.3 we proved that the expectation and variance of the number of occurrences of an event in independent experiments are equal to:
; .
Substituting these expressions into (13.9.8), we get
,
and the formula (13.9.5) takes the form:
.
The theorem is proved.
Example 1. The 100 series of bombs are dropped along the enemy fortifications. When dropping one such series, the expected number of hits is 2, and the standard deviation of the number of hits is 1.5. Approximately find the probability that when dropping 100 episodes from 180 to 220 bombs fall into the lane.
Decision. Imagine the total number of hits as the sum of the numbers of bombs hit in separate series:
,
Where - number of hits th series.
The conditions of the central limit theorem are met, since the values equally distributed. We count the number sufficient in order to be able to apply the limit theorem (in practice it is usually applicable even for much smaller ). We have:
, .
Applying the formula (13.9.6), we get:
,
that is, with a probability of 0.82, it can be argued that the total number of hits in the band will not go beyond .
Example 2. A group air battle occurs in which 50 bombers and 100 fighters participate. Each bomber is attacked by two fighters; Thus, the air combat falls into 50 elementary air battles, each of which involves one bomber and two fighters. In each elementary battle, the probability of shooting down a bomber is 0.4; the probability that both fighters will be shot down in elementary combat is 0.2: the probability that exactly one fighter will be shot down is 0.5. Required: 1) to find the probability that at least 35% of the bombers will be shot down in an air battle; 2) estimate the boundaries in which the number of downed fighters will be concluded with a probability of 0.9.
Decision. 1) denote - The number of downed bombers;
,
Where - the number of bombers shot down m elementary battle.
Distribution range has the form:
From here
; ; ; .
Applying the formula (13.9.6) and assuming (or, equivalently, in this case ), , we find:
.
2) denote the number of downed fighters:
,
Where - the number of fighters shot down in m elementary battle.
Distribution range has the form:
From here we find the mathematical expectation and the variance of :
; .
For size :
; ; .
Determine the boundaries of the plot, symmetric with respect to which the value . Denote half the length of this section. . Then
,
.
By function tables find the value of the argument for which ; this value is approximately equal
,
those.
,
from where
.
Consequently, with a probability of about 0.9, it can be argued that the number of downed fighters will be enclosed in the limit , that is, in the range from 37 to 53.
Comments
To leave a comment
Probability theory. Mathematical Statistics and Stochastic Analysis
Terms: Probability theory. Mathematical Statistics and Stochastic Analysis