Difference between revisions of "Theory of Stochastic Signals/Binomial Distribution"

From LNTwww
Line 11: Line 11:
 
The  '''binomial distribution'''  represents an important special case for the occurrence probabilities of a discrete random variable.
 
The  '''binomial distribution'''  represents an important special case for the occurrence probabilities of a discrete random variable.
  
 
+
To derive the binomial distribution,  we assume that  $I$  binary and statistically independent random variables  $b_i$  each can achieve
To derive the binomial distribution, we assume that  $I$ binary and statistically independent random variables  $b_i$  each can achieve
 
 
*the value  $1$  with probability  ${\rm Pr}(b_i = 1) = p$,  and  
 
*the value  $1$  with probability  ${\rm Pr}(b_i = 1) = p$,  and  
 
*the value   $0$  with probability  ${\rm Pr}(b_i = 0) = 1-p$.  
 
*the value   $0$  with probability  ${\rm Pr}(b_i = 0) = 1-p$.  
  
  
 
+
Then the sum  $z$  is also a discrete random variable with the symbol set   $\{0, \ 1, \ 2,\hspace{0.1cm}\text{ ...} \hspace{0.1cm}, \ I\}$,  which is called binomially distributed:
Then the sum  $z$  is also a discrete random variable with the symbol set   $\{0, \ 1, \ 2,\hspace{0.1cm}\text{ ...} \hspace{0.1cm}, \ I\}$, , which is called binomially distributed:
 
  
 
:$$z=\sum_{i=1}^{I}b_i.$$  
 
:$$z=\sum_{i=1}^{I}b_i.$$  
Thus, the symbol size is  $M = I + 1.$ }}
+
Thus,  the symbol set size is  $M = I + 1.$ }}
  
  
 
{{GraueBox|TEXT=   
 
{{GraueBox|TEXT=   
 
$\text{Example 1:}$ 
 
$\text{Example 1:}$ 
The binomial distribution finds manifold applications in communications engineering as well as in other disciplines:  
+
The binomial distribution finds manifold applications in Communications Engineering as well as in other disciplines:  
 
#  It describes the distribution of rejects in statistical quality control.
 
#  It describes the distribution of rejects in statistical quality control.
 
#  It allows the calculation of the residual error probability in blockwise coding.
 
#  It allows the calculation of the residual error probability in blockwise coding.
# Also the bit error rate of a digital transmission system obtained by simulation is actually a binomially distributed random quantity.
+
# The bit error rate of a digital transmission system obtained by simulation is actually a binomially distributed random quantity.}}
Probabilities of the binomial distribution.}}
 
  
 
==Probabilities of the binomial distribution==
 
==Probabilities of the binomial distribution==
Line 38: Line 35:
 
For the   '''probabilities of the binomial distribution'''   with   $μ = 0, \hspace{0.1cm}\text{...} \hspace{0.1cm}, \ I$:
 
For the   '''probabilities of the binomial distribution'''   with   $μ = 0, \hspace{0.1cm}\text{...} \hspace{0.1cm}, \ I$:
 
:$$p_\mu = {\rm Pr}(z=\mu)={I \choose \mu}\cdot p\hspace{0.05cm}^\mu\cdot ({\rm 1}-p)\hspace{0.05cm}^{I-\mu}.$$
 
:$$p_\mu = {\rm Pr}(z=\mu)={I \choose \mu}\cdot p\hspace{0.05cm}^\mu\cdot ({\rm 1}-p)\hspace{0.05cm}^{I-\mu}.$$
The first term here indicates the number of combinations   $($read:   $I\ \text{  over  }\ μ)$:
+
*The first term here indicates the number of combinations   $($read:   $I\ \text{  over  }\ μ)$:
 
:$${I \choose \mu}=\frac{I !}{\mu !\cdot (I-\mu) !}=\frac{ {I\cdot (I- 1) \cdot \ \cdots \ \cdot (I-\mu+ 1)} }{ 1\cdot  2\cdot \ \cdots \ \cdot  \mu}.$$}}
 
:$${I \choose \mu}=\frac{I !}{\mu !\cdot (I-\mu) !}=\frac{ {I\cdot (I- 1) \cdot \ \cdots \ \cdot (I-\mu+ 1)} }{ 1\cdot  2\cdot \ \cdots \ \cdot  \mu}.$$}}
  
  
''Additional notes:''
+
Additional notes:
*For very large values of   $I$ , the binomial distribution can be approximated by the   [[Theory_of_Stochastic_Signals/Poisson_Distribution|Poisson distribution]]  described in the next section.
+
*For very large values of   $I$,  the binomial distribution can be approximated by the   [[Theory_of_Stochastic_Signals/Poisson_Distribution|Poisson distribution]]  described in the next section.
*If at the same time the product   $I · p \gg 1$, , then according to   [https://en.wikipedia.org/wiki/De_Moivre%E2%80%93Laplace_theorem de Moivre–Laplace's (central limit) theorem]  , the Poisson distribution   (and hence the binomial distribution))  transitions to a discrete  [[Theory_of_Stochastic_Signals/Gaussian_Distributed_Random_Variables|Gaussian distribution]] .
+
*If at the same time the product   $I · p \gg 1$,  then according to   [https://en.wikipedia.org/wiki/De_Moivre%E2%80%93Laplace_theorem de Moivre–Laplace's (central limit) theorem],  the Poisson distribution   (and hence the binomial distribution)  transitions to a discrete  [[Theory_of_Stochastic_Signals/Gaussian_Distributed_Random_Variables|Gaussian distribution]].
  
  
[[File:P_ID203__Sto_T_2_3_S2_neu.png |frame| Probabilites of the binomial distribution]]
 
 
{{GraueBox|TEXT=   
 
{{GraueBox|TEXT=   
 
$\text{Example 2:}$ 
 
$\text{Example 2:}$ 
The graph shows the probabilities of the binomial distribution are for  $I =6$   and  $p =0.4$.  Thus  $M = I+1=7$  probabilities are different from zero.
+
The graph shows the probabilities of the binomial distribution are for  $I =6$   and  $p =0.4$.   
 +
[[File:P_ID203__Sto_T_2_3_S2_neu.png |frame|Binomial distribution probabilites]]
 +
*Thus  $M = I+1=7$  probabilities are different from zero.
  
In contrast, for  $I = 6$  and  $p = 0.5$, the binomial probabilities are as follows:  
+
*In contrast,  for  $I = 6$  and  $p = 0.5$,  the probabilities of the binomial distribution are as follows:  
 
:$$\begin{align*}{\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}0)  & =  {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}6)\hspace{-0.05cm} =\hspace{-0.05cm} 1/64\hspace{-0.05cm} = \hspace{-0.05cm}0.015625 ,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}1)  & =  {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}5) \hspace{-0.05cm}= \hspace{-0.05cm}6/64 \hspace{-0.05cm}=\hspace{-0.05cm} 0.09375,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}2)  & =  {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}4)\hspace{-0.05cm} = \hspace{-0.05cm}15/64 \hspace{-0.05cm}= \hspace{-0.05cm}0.234375 ,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}3)  & =  20/64 \hspace{-0.05cm}= \hspace{-0.05cm} 0.3125 .\end{align*}$$
 
:$$\begin{align*}{\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}0)  & =  {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}6)\hspace{-0.05cm} =\hspace{-0.05cm} 1/64\hspace{-0.05cm} = \hspace{-0.05cm}0.015625 ,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}1)  & =  {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}5) \hspace{-0.05cm}= \hspace{-0.05cm}6/64 \hspace{-0.05cm}=\hspace{-0.05cm} 0.09375,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}2)  & =  {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}4)\hspace{-0.05cm} = \hspace{-0.05cm}15/64 \hspace{-0.05cm}= \hspace{-0.05cm}0.234375 ,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}3)  & =  20/64 \hspace{-0.05cm}= \hspace{-0.05cm} 0.3125 .\end{align*}$$
  
These are symmetrical with respect to the abscissa value   $\mu = I/2 = 3$.}}
+
*These are symmetrical with respect to the abscissa value   $\mu = I/2 = 3$.}}
  
  
Another example of the application of the binomial distribution is the  '''calculation of the block error probability in digital transmission'''.
+
{{GraueBox|TEXT= 
 +
$\text{Example 3:}$  Another example of the application of the binomial distribution is the  '''Calculation of the block error probability in Digital Signal Transmission'''.
  
{{GraueBox|TEXT= 
+
If one transmits blocks each of  $I =10$  binary symbols over a channel
$\text{Example 3:}$ 
+
*with probability  $p = 0.01$  that one symbol is corrupted   ⇒   random variable  $e_i = 1$,  and  
If one transmits blocks of  $I =10$  binary symbols each over a channel which is
+
*correspondingly with probability  $1 - p = 0.99$  for a uncorrupted  symbol   ⇒   random variable  $e_i = 0$,  
*with probability  $p = 0.01$  one symbol is corrupted   ⇒   random variable  $e_i = 1$,  and  
 
*correspondingly with probability  $1 - p = 0.99$  transmits the symbol uncorrupted   ⇒   random variable  $e_i = 0$,  
 
  
  
then the new random variable  $f$  ("block error") is:  
+
then the new random variable  $f$   ⇒   "number of block error"   is:  
 
:$$f=\sum_{i=1}^{I}e_i.$$
 
:$$f=\sum_{i=1}^{I}e_i.$$
  
This random variable  $f$  can now take all integer values between  $0$  (no symbol corrupted)  and  $I$  (all symbols incorrect) .  We denote the probabilities for  $\mu$  corruptions by  $p_μ$.  
+
This random variable  $f$  can take all integer values between  $0$  (no symbol is corrupted)  and  $I$  (all symbols symbol are corrupted) .   
*The case where all  $I$  symbols are correctly transmitted occurs with probability  $p_0 = 0.99^{10} ≈ 0.9044$ . This also follows from the binomial formula for  $μ = 0$  considering definition  $10\, \text{ over }\, 0 = 1$.  
+
*We denote the probabilities for  $\mu$  corruptions by  $p_μ$.  
 +
*The case where all  $I$  symbols are correctly transmitted occurs with probability  $p_0 = 0.99^{10} ≈ 0.9044$ .  
 +
*This also follows from the binomial formula for  $μ = 0$  considering the definition  $10\, \text{ over }\, 0 = 1$.  
 
*A single symbol error  $(f = 1)$  occurs with the following probability:  
 
*A single symbol error  $(f = 1)$  occurs with the following probability:  
 
:$$p_1 = \rm 10\cdot 0.01\cdot 0.99^9\approx 0.0914.$$
 
:$$p_1 = \rm 10\cdot 0.01\cdot 0.99^9\approx 0.0914.$$
:The first factor considers that there are exactly  $10\, \text{ over }\, 1 = 10$  possibilities for the position of a single error.  The other two factors take into account that one symbol must be corrupted and nine must be transmitted correctly if  $f =1$  is to hold.  
+
::The first factor considers that there are exactly  $10\, \text{ over }\, 1 = 10$  possibilities for the position of a single error.   
*For  $f =2$  there are clearly more combinations, namely  $10\, \text{ over }\, 2 = 45$,  and we get  
+
::The other two factors take into account that one symbol must be corrupted and nine must be transmitted correctly if  $f =1$  is to hold.  
 +
*For  $f =2$  there are clearly more combinations,  namely  $10\, \text{ over }\, 2 = 45$,   and we get  
 
:$$p_2 = \rm 45\cdot 0.01^2\cdot 0.99^8\approx 0.0041.$$
 
:$$p_2 = \rm 45\cdot 0.01^2\cdot 0.99^8\approx 0.0041.$$
  
If a block code can correct up to two errors, the residual error probability is
+
If a block code can correct up to two errors,  the block error probability is
:$$p_{\rm R} = \it p_{\rm 3} \rm +\hspace{0.1cm}\text{ ...} \hspace{0.1cm} \rm + \it p_{\rm 10}\approx \rm 10^{-4},$$
+
:$$p_{\rm block} = \it p_{\rm 3} \rm +\hspace{0.1cm}\text{ ...} \hspace{0.1cm} \rm + \it p_{\rm 10}\approx \rm 10^{-4},$$
oder
+
or
:$$p_{\rm R} = \rm 1-\it p_{\rm 0}-\it p_{\rm 1}-p_{\rm 2}\approx \rm 10^{-4}.$$
+
:$$p_{\rm block} = \rm 1-\it p_{\rm 0}-\it p_{\rm 1}-p_{\rm 2}\approx \rm 10^{-4}.$$
  
*One can see that the second possibility of calculation via the complement for large values of  $I$  leads faster to the goal.  
+
*One can see that for large values of  $I$  the second possibility of calculation via the complement leads faster to the goal.  
*However, one could also consider as an approximation that for these numerical values  $p_{\rm R} ≈ p_3$  holds. }}
+
*However,   one could also consider that for these numerical values  $p_{\rm block} ≈ p_3$  holds as an approximation. }}
  
  
Use the interactive applet  [[Applets:Binomial-_und_Poissonverteilung_(Applet)|Binomial and Poisson distribution]]  to find the binomial probabilities for any  $I$  and  $p$ .
+
Use the interactive HTML5/JS applet  [[Applets:Binomial-_und_Poissonverteilung_(Applet)|Binomial and Poisson distribution]]  to find the binomial probabilities for any  $I$  and  $p$ .
  
  

Revision as of 11:01, 11 December 2021

General description of the binomial distribution


$\text{Definition:}$  The  binomial distribution  represents an important special case for the occurrence probabilities of a discrete random variable.

To derive the binomial distribution,  we assume that  $I$  binary and statistically independent random variables  $b_i$  each can achieve

  • the value  $1$  with probability  ${\rm Pr}(b_i = 1) = p$,  and
  • the value  $0$  with probability  ${\rm Pr}(b_i = 0) = 1-p$.


Then the sum  $z$  is also a discrete random variable with the symbol set   $\{0, \ 1, \ 2,\hspace{0.1cm}\text{ ...} \hspace{0.1cm}, \ I\}$,  which is called binomially distributed:

$$z=\sum_{i=1}^{I}b_i.$$

Thus,  the symbol set size is  $M = I + 1.$


$\text{Example 1:}$  The binomial distribution finds manifold applications in Communications Engineering as well as in other disciplines:

  1.   It describes the distribution of rejects in statistical quality control.
  2.   It allows the calculation of the residual error probability in blockwise coding.
  3.  The bit error rate of a digital transmission system obtained by simulation is actually a binomially distributed random quantity.

Probabilities of the binomial distribution


$\text{Calculation rule:}$  For the  probabilities of the binomial distribution  with  $μ = 0, \hspace{0.1cm}\text{...} \hspace{0.1cm}, \ I$:

$$p_\mu = {\rm Pr}(z=\mu)={I \choose \mu}\cdot p\hspace{0.05cm}^\mu\cdot ({\rm 1}-p)\hspace{0.05cm}^{I-\mu}.$$
  • The first term here indicates the number of combinations   $($read:  $I\ \text{ over }\ μ)$:
$${I \choose \mu}=\frac{I !}{\mu !\cdot (I-\mu) !}=\frac{ {I\cdot (I- 1) \cdot \ \cdots \ \cdot (I-\mu+ 1)} }{ 1\cdot 2\cdot \ \cdots \ \cdot \mu}.$$


Additional notes:


$\text{Example 2:}$  The graph shows the probabilities of the binomial distribution are for  $I =6$  and  $p =0.4$. 

Binomial distribution probabilites
  • Thus  $M = I+1=7$  probabilities are different from zero.
  • In contrast,  for  $I = 6$  and  $p = 0.5$,  the probabilities of the binomial distribution are as follows:
$$\begin{align*}{\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}0) & = {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}6)\hspace{-0.05cm} =\hspace{-0.05cm} 1/64\hspace{-0.05cm} = \hspace{-0.05cm}0.015625 ,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}1) & = {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}5) \hspace{-0.05cm}= \hspace{-0.05cm}6/64 \hspace{-0.05cm}=\hspace{-0.05cm} 0.09375,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}2) & = {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}4)\hspace{-0.05cm} = \hspace{-0.05cm}15/64 \hspace{-0.05cm}= \hspace{-0.05cm}0.234375 ,\\ {\rm Pr}(z\hspace{-0.05cm} =\hspace{-0.05cm}3) & = 20/64 \hspace{-0.05cm}= \hspace{-0.05cm} 0.3125 .\end{align*}$$
  • These are symmetrical with respect to the abscissa value  $\mu = I/2 = 3$.


$\text{Example 3:}$  Another example of the application of the binomial distribution is the  Calculation of the block error probability in Digital Signal Transmission.

If one transmits blocks each of  $I =10$  binary symbols over a channel

  • with probability  $p = 0.01$  that one symbol is corrupted   ⇒   random variable  $e_i = 1$,  and
  • correspondingly with probability  $1 - p = 0.99$  for a uncorrupted symbol   ⇒   random variable  $e_i = 0$,


then the new random variable  $f$   ⇒   "number of block error"  is:

$$f=\sum_{i=1}^{I}e_i.$$

This random variable  $f$  can take all integer values between  $0$  (no symbol is corrupted)  and  $I$  (all symbols symbol are corrupted) . 

  • We denote the probabilities for  $\mu$  corruptions by  $p_μ$.
  • The case where all  $I$  symbols are correctly transmitted occurs with probability  $p_0 = 0.99^{10} ≈ 0.9044$ .
  • This also follows from the binomial formula for  $μ = 0$  considering the definition  $10\, \text{ over }\, 0 = 1$.
  • A single symbol error  $(f = 1)$  occurs with the following probability:
$$p_1 = \rm 10\cdot 0.01\cdot 0.99^9\approx 0.0914.$$
The first factor considers that there are exactly  $10\, \text{ over }\, 1 = 10$  possibilities for the position of a single error. 
The other two factors take into account that one symbol must be corrupted and nine must be transmitted correctly if  $f =1$  is to hold.
  • For  $f =2$  there are clearly more combinations,  namely  $10\, \text{ over }\, 2 = 45$,   and we get
$$p_2 = \rm 45\cdot 0.01^2\cdot 0.99^8\approx 0.0041.$$

If a block code can correct up to two errors,  the block error probability is

$$p_{\rm block} = \it p_{\rm 3} \rm +\hspace{0.1cm}\text{ ...} \hspace{0.1cm} \rm + \it p_{\rm 10}\approx \rm 10^{-4},$$

or

$$p_{\rm block} = \rm 1-\it p_{\rm 0}-\it p_{\rm 1}-p_{\rm 2}\approx \rm 10^{-4}.$$
  • One can see that for large values of  $I$  the second possibility of calculation via the complement leads faster to the goal.
  • However,  one could also consider that for these numerical values  $p_{\rm block} ≈ p_3$  holds as an approximation.


Use the interactive HTML5/JS applet  Binomial and Poisson distribution  to find the binomial probabilities for any  $I$  and  $p$ .


Moments of the binomial distribution


You can calculate the moments in general using the equations in the chapters  Moments of a discrete random variable  and  probabilities of the binomial distribution .

$\text{Calculation rules:} $  For the  $k$-th order moment  of a binomially distributed random variable, the general rule is:

$$m_k={\rm E}\big[z^k\big]=\sum_{\mu={\rm 0} }^{I}\mu^k\cdot{I \choose \mu}\cdot p\hspace{0.05cm}^\mu\cdot ({\rm 1}-p)\hspace{0.05cm}^{I-\mu}.$$

From this, after some transformations, we obtain for

  • the linear mean value:
$$m_1 ={\rm E}\big[z\big]= I\cdot p,$$
  • the rms value:
$$m_2 ={\rm E}\big[z^2\big]= (I^2-I)\cdot p^2+I\cdot p.$$

The variance and standard deviation are obtained by applyings "Steiner's theorem":

$$\sigma^2 = {m_2-m_1^2} = {I \cdot p\cdot (1-p)} \hspace{0.3cm}\Rightarrow \hspace{0.3cm} \sigma = \sqrt{I \cdot p\cdot (1-p)}.$$


The maximum variance  $σ^2 = I/4$  is obtained for the "characteristic probability"  $p = 1/2$.  In this case, the probabilities are symmetric around the mean  $m_1 = I/2 \ ⇒ \ p_μ = p_{I–μ}$.

The more the characteristic probability  $p$  deviates from the value  $1/2$ ,

  • the smaller is the standard deviation  $σ$, and
  • the more asymmetric the probabilities become around the mean  $m_1 = I · p$.


$\text{Example 4:}$  As in  $\text{example 3}$ , we consider a block of  $I =10$  binary symbols, each of which is independently corrupted with probability  $p = 0.01$ .  Then holds:

  • The mean number of block errors is equal to  $m_f = {\rm E}\big[ f\big] = I · p = 0.1$.
  • Der standard deviation of the random variable  $f$  is  $σ_f = \sqrt{0.1 \cdot 0.99}≈ 0.315$.


In contrast, in the completely corrupted channel   ⇒   falsification probability (*besseres Wort als falsification für Verfälschung?*)  $p = 1/2$  results in the values

  • $m_f = 5$   ⇒   on average, five of the ten bits within a block are wrong,
  • $σ_f = \sqrt{I}/2 ≈1.581$   ⇒   maximum standard deviation  $I = 10$.

Exercises for the chapter


Exercise 2.3: Algebraic Sum of Binary Numbers

Exercise 2.4: Number Lottery (6 from 49)