์ด๋ฒ ๊ธ์์๋ 2๊ฐ ์ด์์ ํ๋ฅ ๋ณ์(๋ผ๊ณ ๋ ํ์ง๋ง ๋ชจ๋ ์ค๋ช ์ 2๊ฐ์ผ ๋๋ฅผ ๊ฐ์ง๊ณ ์ค๋ช ํฉ๋๋ค)์ ์กฐ๊ฑด๋ถ ํ๋ฅ ๊ณผ ๊ด๋ จํ์ฌ ์์๋ณด๋๋ก ํ๊ฒ ์ต๋๋ค.
joint p.f ํน์ p.d.f๋ฅผ ๊ฐ์ง๋ (์ด์ฐ ํน์ ์ฐ์)ํ๋ฅ ๋ณ์ X, Y์ ๋ํด,
$$f_1 \;:\;marginal \;\; p.f \; / \; p.d.f \;\; of \;\; X $$
$$f_2 \;:\;marginal \;\; p.f \; / \; p.d.f \;\; of \;\; Y $$
$$g_1(x|y) = \frac{f(x,y)}{f_2(y)} \;\;\; for \;\; \forall_{Y} \;\; f_2(y)>0$$
Discreteํ ๊ฒฝ์ฐ์ ๋ํ ์ฆ๋ช
$$f(x | y) = P_{X|Y}(X=x | Y=y) = \frac{P(X = x, Y = y)}{P(Y=Y)}$$
$$๋ฐ๋ผ์ \;\;g_1(x|y) = \frac{f(x,y)}{f_2(y)} \;\;\; for \;\; \forall_{Y} \;\; f_2(y)>0$$
Continuousํ ๊ฒฝ์ฐ
Conditional Probability๋ฅผ Joint Probability์ ์ ์ฅ์์ ๋ณธ๋ค๋ฉด ๋ค์๊ณผ ๊ฐ์ด ํํํฉ๋๋ค.
$$f(y|x) = f(y |X=x)$$
์ด์ ์ด ๊ฒฝ์ฐ์์์ ํ๋ฅ ๋ถํฌ, ์ฆ ์กฐ๊ฑด๋ถ ํ๋ฅ ๋ฐ๋ํจ์๋ฅผ ๊ตฌํด๋ณด๊ฒ ์ต๋๋ค.
$$f(y|x)=\frac{\partial }{\partial y}F(y | X =x)=\frac{\partial }{\partial y}P(Y \leq y| X = x)$$
$$=\frac{\partial }{\partial y}\frac{P(y \leq Y, X =x)}{P(X=x)}$$
๊ทธ๋ฐ๋ฐ X์ Y๋ Continuous RV์ด๊ณ , ๋ฐ๋ผ์ X = x ์ผ ํ๋ฅ ์ 0์ ์๋ ดํฉ๋๋ค. ์ฆ ๋ฐ๋ก ์ ์์์ ๋ถ์์ธ P(Y <= y , X = x)์ ๋ถ๋ชจ์ธ P(X = x)์ ๋ชจ๋ 0์ผ๋ก ์๋ ดํ๊ธฐ์ ๋ค์๊ณผ ๊ฐ์ด ์์ฑํ ์ ์์ต๋๋ค.
$$\frac{\partial }{\partial y}\displaystyle \lim_{\Delta x \to 0}\frac{P(Y \leq y, x < X \leq x +\Delta x)}{P(x < X \leq x + \Delta x)}$$
์ด๋ CDF๋ก ํํํ๋ฉด ๋ค์๊ณผ ๊ฐ์์ง๋๋ค.
$$\frac{\partial }{\partial y}\displaystyle \lim_{\Delta x \to 0}\frac{F_{xy}(x + \Delta x, \; y) - F_{xy}(x,y)}{F_x(x + \Delta x) - F_x(x)}$$
๋ถ๋ชจ์ ๋ถ์์ ๋ฏธ๋ถ์ ํ์์ ์ทจํ๋ฉด
$$\frac{\partial }{\partial y}\displaystyle \lim_{\Delta x \to 0}\frac{[F_{xy}(x + \Delta x, \; y) - F_{xy}(x,y)]/\Delta x}{[F_x(x + \Delta x) - F_x(x)]/\Delta x}$$
$$ = \frac{\partial }{\partial y}\frac{\frac{\partial }{\partial x}F_{xy}(x,y)}{f_x(x)}$$
$$ = \frac{\frac{\partial^{2} }{\partial x\partial y}F_{xy}(x,y)}{f_x(x)}$$
$$=\frac{f_{xy}(x, y)}{f_x(x)}$$
๋ฐ๋ผ์ ์ ๋ฆฌํ๋ฉด ๋ค์๊ณผ ๊ฐ์ต๋๋ค.
$$f(y | x) =\frac{f(x, y)}{f_x(x)} $$
๋ํ X์ Y๊ฐ Independentํ ๊ฒฝ์ฐ๋ ์๋ ์์ด ์ฑ๋ฆฝํ ๋์ ๋๋ค.
$$f(x|y) = f_x(x)$$