2021-1ν•™κΈ°, λŒ€ν•™μ—μ„œ β€˜ν†΅κ³„μ  λ°μ΄ν„°λ§ˆμ΄λ‹β€™ μˆ˜μ—…μ„ λ“£κ³  κ³΅λΆ€ν•œ λ°”λ₯Ό μ •λ¦¬ν•œ κΈ€μž…λ‹ˆλ‹€. 지적은 μ–Έμ œλ‚˜ ν™˜μ˜μž…λ‹ˆλ‹€ :)

1 minute read

2021-1ν•™κΈ°, λŒ€ν•™μ—μ„œ β€˜ν†΅κ³„μ  λ°μ΄ν„°λ§ˆμ΄λ‹β€™ μˆ˜μ—…μ„ λ“£κ³  κ³΅λΆ€ν•œ λ°”λ₯Ό μ •λ¦¬ν•œ κΈ€μž…λ‹ˆλ‹€. 지적은 μ–Έμ œλ‚˜ ν™˜μ˜μž…λ‹ˆλ‹€ :)

Motivation.

estimatorμ—μ„œ independent vector $\mathbf{x}$μ—μ„œ μ–΄λ–€ featureκ°€ response vector $\mathbf{y}$에 영ν–₯을 λ―ΈμΉ˜λŠ”μ§€ ν™•μΈν•˜λ €λ©΄ μ–΄λ–»κ²Œ ν•΄μ•Όν• κΉŒ? κ°„λ‹¨ν•˜κ²Œ 생각해본닀면, μΆ”μ •ν•œ $\hat{\beta}$μ—μ„œ $\hat{\beta}_i$의 값이 0인지 μ•„λ‹Œμ§€λ₯Ό ν†΅ν•΄μ„œ νŒλ‹¨ν•  수 μžˆμ„ 것이닀. μ΄λ ‡κ²Œ μ–΄λ–€ featureκ°€ 결과에 영ν–₯을 λ―ΈμΉœλ‹€ μ•ˆ λ―ΈμΉœλ‹€λ₯Ό μ°Ύμ•„λ‚΄λŠ” μž‘μ—…μ„ <톡계적 μΆ”λ‘  statistical inference>라고 ν•œλ‹€.

μ•„λž˜μ˜ 가정은 <statistical inference>λ₯Ό μˆ˜ν–‰ν•  λ•Œμ— μ‹œν–‰ν•˜λŠ” 고전적인 가정이닀.

Assumption. Classical Assumption

Assume that the true distribution of the data is

\[Y = X^T \beta + \epsilon, \quad \epsilon \sim N(0, \sigma^2)\]

이것을 λ‹€μ‹œ μ“°λ©΄,

\[(Y \mid X = x) \sim N(x^T \beta, \; \sigma^2)\]

λ§Œμ•½ μœ„μ™€ 같은 가정을 λ§Œμ‘±ν•œλ‹€λ©΄, μ•„λž˜μ˜ μ„±μ§ˆμ΄ 성립함을 증λͺ…ν•  수 μžˆλ‹€.

Property.

Supp. that the classical assumption holds. Then,

\[\hat{\beta} \sim N(\beta, \; (\mathbf{X}^T \mathbf{X})^{-1} \sigma^2)\]

그리고 $\hat{\sigma}^2$λ₯Ό μ λ‹Ήνžˆ scaling ν•΄μ€€λ‹€λ©΄,

\[\frac{(n-p) \hat{\sigma}^2}{\sigma^2} \sim \chi^2_{n-p}\]

그리고, $\hat{\beta}$, $\hat{\sigma}^2$λŠ” μ„œλ‘œ independentν•˜λ‹€.

\[\hat{\beta} \perp \hat{\sigma}^2\]

이 뢀뢄은 좔후에 쒀더 λ³΄μΆ©ν•˜λ„λ‘ ν•˜κ² λ‹€.