โ€œํ™•๋ฅ ๊ณผ ํ†ต๊ณ„(MATH230)โ€ ์ˆ˜์—…์—์„œ ๋ฐฐ์šด ๊ฒƒ๊ณผ ๊ณต๋ถ€ํ•œ ๊ฒƒ์„ ์ •๋ฆฌํ•œ ํฌ์ŠคํŠธ์ž…๋‹ˆ๋‹ค. ์ „์ฒด ํฌ์ŠคํŠธ๋Š” Probability and Statistics์—์„œ ํ™•์ธํ•˜์‹ค ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค ๐ŸŽฒ

5 minute read

โ€œํ™•๋ฅ ๊ณผ ํ†ต๊ณ„(MATH230)โ€ ์ˆ˜์—…์—์„œ ๋ฐฐ์šด ๊ฒƒ๊ณผ ๊ณต๋ถ€ํ•œ ๊ฒƒ์„ ์ •๋ฆฌํ•œ ํฌ์ŠคํŠธ์ž…๋‹ˆ๋‹ค. ์ „์ฒด ํฌ์ŠคํŠธ๋Š” Probability and Statistics์—์„œ ํ™•์ธํ•˜์‹ค ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค ๐ŸŽฒ

Weibull Distribution

Definition.

Let $\alpha > 0$ and $\beta > 0$. We say that a RV $X$ has a <Weibull distribution>, denoted as $X \sim \text{Weibull}(\alpha, \beta)$, if its pdf $f(x)$ is given by

\[f(x; \alpha, \beta) = \alpha \beta \cdot x^{\beta - 1} \cdot e^{-\alpha x^{\beta}} \quad \text{for } x > 0\]

<Weibull Distribution>์€ ์ด๋Ÿฐ ๋ถ„ํฌ๊ฐ€ ์žˆ๋‹ค ์ •๋„๋งŒ ์†Œ๊ฐœํ•˜๊ณ  ๋„˜์–ด๊ฐ„๋‹ค.

Remark.

1. Relationship with Exponential Distribution

if $\beta = 1$, then $\text{Weibull}(\alpha, 1) = \text{Exp}(\alpha)$.


2. cdf of $X$ is

\[F(x) = \int^x_0 f(y) \, dy = \begin{cases} 1 - e^{-\alpha x^{\beta}} & \text{for } x > 0 \\ \quad 0 & \text{else} \end{cases}\]

์œ„์˜ ์‹์„ ๋ฏธ๋ถ„ํ•ด๋ณด๋ฉด, Weibull์˜ pdf๊ฐ€ ๋‚˜์˜จ๋‹ค๋Š” ๊ฑธ ์‰ฝ๊ฒŒ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค.


Failure rate & Reliability

Let $T$ be a RV representing the lifetime (or time to failure) of a certain component.

Let $f(t)$ and $F(t)$ be its pdf and cdf respectively.

Definition.

1. reliability function, or survival function

\[R(t) := P(T > t) = 1 - F(t)\]

์ฆ‰, CDF์˜ tail probability๋‹ค. ์™œ๋ƒํ•˜๋ฉด, $P(T > t)$๋Š” component๊ฐ€ $[0, t]$ ๋™์•ˆ surviveํ•  ํ™•๋ฅ ์„ ์˜๋ฏธํ•˜๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค!

2. failure rate, or hazard rate

\[Z(t) := \frac{f(t)}{R(t)}\]

Q. Why?

\[\begin{aligned} f(t) &= \frac{d}{dt} F(t) \\ &= \lim_{h \rightarrow 0} \frac{F(t+h) - F(t)}{h} \\ &= \lim_{h \rightarrow 0} \frac{P(T < t+h) - P(T < t)}{h} \\ &= \lim_{h \rightarrow 0} \frac{P(t < T \le t+h)}{h} \end{aligned}\]

์ด๋•Œ, ์œ„์˜ ์‹์— $R(t)$๋ฅผ ๋‚˜๋ˆ ๋ณด์ž!

\[\begin{aligned} \frac{f(t)}{R(t)} &= \lim_{h \rightarrow 0} \frac{P(t < T \le t+h)}{h} \cdot \frac{1}{R(t)} \\ &= \lim_{h \rightarrow 0} \frac{P(t < T \le t+h)}{h \cdot R(t)} \\ &= \lim_{h \rightarrow 0} \frac{1}{h} \cdot \frac{P(t < T \le t+h)}{P(T > t)} \end{aligned}\]

์œ„์˜ ์‹์—์„œ ๋ณผ ์ˆ˜ ์žˆ๋“ฏ, condition probability $\dfrac{P(t < T \le t+h)}{P(t > t)} = P(t < T \le t+h \mid T > t )$์ด ๋œ๋‹ค. ๊ทธ๋ž˜์„œ ์‹์„ ์ •๋ฆฌํ•˜๋ฉด,

\[\frac{f(t)}{R(t)} = \lim_{h \rightarrow 0} \frac{P(t < T \le t+h \mid T > t)}{h}\]

์œ„์˜ ์‹์€ failure rate $f(t)/R(t)$๊ฐ€ โ€œthe rate of the probability of the failure right after time $t$โ€์ž„์„ ์˜๋ฏธํ•œ๋‹ค! $\blacksquare$

๋งŒ์•ฝ $T$๊ฐ€ Weibull distribution์„ ๋”ฐ๋ฅธ๋‹ค๋ฉด, failure rate $Z(t)$๋Š”

\[Z(t) = \frac{f(t)}{R(t)} = \frac{\alpha \beta \cdot t^{\beta-1} e^{-\alpha t^{\beta}}}{e^{-\alpha e^{\beta}}} = \alpha \beta \cdot t^{\beta - 1}\]

์ด๋•Œ $\beta$์˜ ๊ฐ’์— ๋”ฐ๋ผ์„œ failure rate์˜ ์–‘์ƒ์„ ์‚ดํŽด๋ณผ ์ˆ˜๋„ ์žˆ๋Š”๋ฐ,

1. if $\beta = 1$, then the failure rate is $\alpha$ (constant).

์ฆ‰, ์‹œ๊ฐ„์— ๊ด€๊ณ„์—†์ด failure rate๋Š” ํ•ญ์ƒ ๊ฐ™๋‹ค.


2. if $\beta > 1$, then failure rate is increasing as $t$ flows.

์ฆ‰, ์‹œ๊ฐ„์ด ์ง€๋‚ ์ˆ˜๋ก ์žฅ๋น„๊ฐ€ ์•ฝํ•ด์ง„๋‹ค๋Š” ๊ฒƒ์„ ์˜๋ฏธํ•œ๋‹ค.


3. if $\beta < 1$, then failure rate if decreasing.

์ฆ‰, ์‹œ๊ฐ„์ด ์ง€๋‚ ์ˆ˜๋ก ์žฅ๋น„๊ฐ€ ์˜คํžˆ๋ ค ๋” ์ข‹์•„์ง€๋Š” ๊ฒƒ์„ ์˜๋ฏธํ•œ๋‹ค.


์ด์–ด์ง€๋Š” ํฌ์ŠคํŠธ์—์„œ๋Š” Random Variable์— ๊ฐ„๋‹จํ•œ ๋ณ€ํ™˜(Transform)์„ ์ ์šฉํ–ˆ์„ ๋•Œ์˜ pdf๋ฅผ ์–ด๋–ป๊ฒŒ ๊ตฌํ•˜๋Š”์ง€ ์‚ดํŽด๋ณธ๋‹ค. ๋’ท๋ถ€๋ถ„์—๋Š” moment์„ ๊ตฌํ•˜๋Š” ํ•จ์ˆ˜์ธ <MGF; Momentim Generating Function>๋„ ๋“ฑ์žฅํ•˜๊ธฐ ๋•Œ๋ฌธ์— ์ค‘์š”ํ•œ ์ฑ•ํ„ฐ๋ผ๊ณ  ํ•  ์ˆ˜ ์žˆ๋‹ค!

๐Ÿ‘‰ Transformations of Random Variable - 1