Since the [[Binomial Distribution]] can be viewed as the sample sum of repeated [[Bernoulli Distribution|Bernoulli]] [[Random Variable|random variables]], it satisfied the assumptions of the [[Central Limit Theorem]].
Thus, a r.v. $X\sim \mathrm{Binom}(n,p)$ can be approximated by a [[Gaussian Distribution]], after being standardized.
**Binomial Distribution:**
$ p_X(X=k) =\sum_{k=0}^{n} \binom{n}{k} * (1-p)^{n-k}* p^n $
**Approximation as Standard Gaussian:** From the properties of the Binomial, we know the [[Expectation]] and [[Variance]], to replace $\mu$ and $\sigma$.
$
Z_n=\frac{X -\mu}{\sigma}
\xrightarrow[n \to \infty]{(d)} \mathcal N(0,1) \quad \text{where}
\begin{cases} \mu=np\\ \sigma^2= np*(1-p)
\end{cases}
$
In terms of binomial parameters:
$ Z_n=\frac{X-np}{\sqrt{np*(1-p)}} \xrightarrow[n \to \infty]{(d)} \mathcal N(0,1) $
## De Moivre-Laplace Correction
While both the Binomial PMF and the Gaussian approximation yield similar results for large $n$, we need to correct for the fact that the latter is a continuous distribution.
Therefore we sum integrate over the interval $[k-0.5, k+0.5]$ to get the probability mass $p_X(X=k)$ at a specific $k$.
**Example:**
Assess the probability of $k=19$, when $X \sim \mathrm{Binom}(n=36; p=0.5)$ via Gaussian approximation.
![[de-moivre-laplace.png|center|400]]
$
\begin{aligned}
\mathbf P(k=19) &= \mathbf P(18.5 \le S_n \le 19.5) \\[8pt]
&=\mathbf P\Big(\frac{18.5-18}{3} \le Z_n \le \frac{19.5-18}{3} \Big) \\[8pt]
&=\mathbf P(0.17 \le Z_n \le 0.5) \\[8pt]
&=\Phi(0.5) - \Phi(0.17) \\[8pt]
&=0.124
\end{aligned}
$