์ผ | ์ | ํ | ์ | ๋ชฉ | ๊ธ | ํ |
---|---|---|---|---|---|---|
1 | 2 | 3 | ||||
4 | 5 | 6 | 7 | 8 | 9 | 10 |
11 | 12 | 13 | 14 | 15 | 16 | 17 |
18 | 19 | 20 | 21 | 22 | 23 | 24 |
25 | 26 | 27 | 28 | 29 | 30 | 31 |
- reducible
- ํญํด99
- 99ํด๋ฝ
- IEEE 802
- tcp ํ๋กํ ์ฝ
- ๋ฐ์ดํฐ ์ ์ก
- mariadb
- xv6
- ๊ทธ๋ฆฌ๋ ์๊ณ ๋ฆฌ์ฆ
- ์ฃผ๊ธฐ์ ํธ
- ๋น์ฃผ๊ธฐ์ ํธ
- ์ฝ๋ฉํ ์คํธ์ค๋น
- ์๋น์ค ํ๋ฆฌ๋ฏธํฐ๋ธ
- leetcode
- ๊ฐ๋ฐ์์ทจ์
- til
- ์ฐ๋ถํฌdb
- well known ํฌํธ
- ์ค๋ฅ์ ์ด
- tcp ์ธ๊ทธ๋จผํธ
- ์์๋ฒํธ
- ์ค๋ฅ๊ฒ์ถ
- ์ค๋ ๋
- i-type
- ํ ํฐ ๋ฒ์ค
- ํ๋ ์ ๊ตฌ์กฐ
- ์ค๋ธ์
- ํ๋ก์ด๋์์
- ํฐ์คํ ๋ฆฌ์ฑ๋ฆฐ์ง
- git merge
- Today
- Total
Unfazedโ๏ธ๐ฏ
F-ํต๊ณ๋ (๊ท๋ฌด ๊ฐ์ค H0 ๊ธฐ๊ฐ ๊ฐ๋ฅ์ฑ) ๋ณธ๋ฌธ
๊ท๋ฌด ๊ฐ์ค H0 ๊ณผ F-ํต๊ณ๋: ๊ท๋ฌด ๊ฐ์ค H0 ์ ํ๊ท ๋ชจ๋ธ์์ ๋ชจ๋ ๋
๋ฆฝ ๋ณ์๊ฐ ์ข
์ ๋ณ์์ ์ํฅ์ ๋ฏธ์น์ง ์๋๋ค๋ ๊ฐ์ค์ด๋ค. ์ด๋ฅผ ๊ฒ์ ํ๊ธฐ ์ํด F-ํต๊ณ๋์ ์ฌ์ฉํ๋ฉฐ, ์ด ํต๊ณ๋์ ๋ชจ๋ธ์ ์ค๋ช
ํ๋ฅ ์ ์ธก์ ํ๋ค.
F-๋ถํฌ: F-ํต๊ณ๋์ ์ ๊ท ๋ถํฌ๋ฅผ ๋ฐ๋ฅด๋ ์ค์ฐจํญ์ ๊ฐ์ ํ ๋ F-๋ถํฌ๋ฅผ ๋ฐ๋ฅธ๋ค. F-๋ถํฌ๋ ๋ ๊ฐ์ ์๋ก ๋ค๋ฅธ ์์ ๋๋ฅผ ๊ฐ์ง๋ฉฐ, ํ๋๋ ํ๊ท์์ ์ฌ์ฉ๋ ๋
๋ฆฝ ๋ณ์์ ์ p์ ์ํด, ๋ค๋ฅธ ํ๋๋ ํ๋ณธ ํฌ๊ธฐ n์์ ๋
๋ฆฝ ๋ณ์์ ์๋ฅผ ๋บ ๊ฐ์ ์ํด ๊ฒฐ์ ๋๋ค.
p-๊ฐ ๊ณ์ฐ: ํต๊ณ ์ํํธ์จ์ด๋ ์ฃผ์ด์ง F-ํต๊ณ๋ ๊ฐ๊ณผ ์์ ๋๋ฅผ ์ฌ์ฉํ์ฌ p-๊ฐ์ ๊ณ์ฐํ๋ค. p-๊ฐ์ ํน์ ํ F-ํต๊ณ๋ ๊ฐ์ด ๊ด์ธก๋ ํ๋ฅ ์ ๋ํ๋ด๋ฉฐ, ์ด ๊ฐ์ด ๋งค์ฐ ๋ฎ๋ค๋ฉด (์ผ๋ฐ์ ์ผ๋ก 0.05 ๋๋ 0.01 ๋ฏธ๋ง) ์ฐ์ฐํ ๋ฐ์ํ๋ค๊ณ ๋ณด๊ธฐ ์ด๋ ค์ฐ๋ฏ๋ก ๊ท๋ฌด ๊ฐ์ค์ ๊ธฐ๊ฐํ๋ค.
๊ฒฐ์ : p-๊ฐ์ด ์ ์ ์์ค(๋ณดํต 0.05)๋ณด๋ค ๋ฎ์ ๊ฒฝ์ฐ, ๊ท๋ฌด ๊ฐ์ค H0 ์ ๊ธฐ๊ฐํ๊ณ , ๋ชจ๋ธ์ ํฌํจ๋ ์ ์ด๋ ํ๋ ์ด์์ ๋
๋ฆฝ ๋ณ์๊ฐ ์ข
์ ๋ณ์์ ์ ์๋ฏธํ ์ํฅ์ ๋ฏธ์น๋ค๊ณ ๊ฒฐ๋ก ์ง์ ์ ์๋ค.
ISLP 84p.
However, what if the F-statistic had been closer to 1? How large does the F-statistic need to be before we can reject H0 and conclude that there is a relationship? It turns out that the answer depends on the values of n and p. When n is large, an F-statistic that is just a little larger than 1 might still provide evidence against H0. In contrast, a larger F-statistic is needed to reject H0 if n is small.
ํ๋ณธ ํฌ๊ธฐ n | F-ํต๊ณ๋ (F-statistic) | ๊ท๋ฌด ๊ฐ์ค H0 ๊ธฐ๊ฐ ๊ฐ๋ฅ์ฑ (Likelihood of Rejecting) |
Large n | Slightly above 1 | (Possible) still provide evidence against H0 |
Small n | Slightly above 1 | Unlikely |
Large n | Large | Very Likely |
Small n | Large | Likely |
When H0 is true and the errors6εi have a normal distribution, the F-statistic follows an F-distribution. For any given value of n and p, any statistical software package can be used to compute the p-value associated with the F -statistic using this distribution. Based on this p-value, we can determine whether or not to reject H0.
p=100์ด๋ฉฐ ๊ท๋ฌด ๊ฐ์ค H0 :β1=β2=โฏ=βp =0์ด ์ฐธ์ด๋ผ๊ณ ๊ฐ์ ํ์. ์ฆ, ์ด๋ค ๋ณ์๋ ๋ฐ์ ๋ณ์์ ์ค์ ๋ก ๊ด๋ จ์ด ์๋ค๊ณ ๊ฐ์ ํ๋ ๊ฒ์ด๋ค. ์ด๋ฌํ ์ํฉ์์ ๊ฐ ๋ณ์์ ๋ํด ๊ณ์ฐ๋ p-๊ฐ ์ค ์ฝ 5%๊ฐ ์ฐ์ฐํ 0.05 ์ดํ๊ฐ ๋ ๊ฒ์ผ๋ก ์์๋๋ค. ๋ค์ ๋งํด, ์ค์ ๋ก ์ด๋ค ๋ณ์์ ๋ฐ์ ๋ณ์ ์ฌ์ด์ ์ฐ๊ด์ฑ์ด ์์์๋ ๋ถ๊ตฌํ๊ณ ์ฝ 5๊ฐ์ ๋ณ์์์๋ p-๊ฐ์ด 0.05 ์ดํ๋ก ๋ํ๋ ์ ์๋ค๋ ๊ฒ์ด๋ค.
ํ์์ ๊ฐ ๊ฐ๋ณ ์์ธก ๋ณ์์ ๋ํด t-ํต๊ณ๋๊ณผ p-๊ฐ์ด ์ ๊ณต๋๋ค ํ์. ์ด ๊ฐ๋ค์ ๋ค๋ฅธ ์์ธก ๋ณ์๋ค์ ์กฐ์ ํ ํ ๊ฐ ๊ฐ๋ณ ์์ธก ๋ณ์๊ฐ ๋ฐ์ ๋ณ์์ ๊ด๋ จ์ด ์๋์ง์ ๋ํ ์ ๋ณด๋ฅผ ์ ๊ณตํ๋ค. ์ด๋ ๋ชจ๋ธ์์ ๋จ์ผ ๋ณ์๋ฅผ ์ ์ธํ๊ณ ๋๋จธ์ง ๋ณ์๋ค์ ํฌํจ์์ผ ์ค์ํ F-๊ฒ์ ๊ณผ ๋๋ฑํ ์ ๋ณด๋ฅผ ์ ๊ณตํ๋ค(์ฆ, q=1์ธ ๊ฒฝ์ฐ). ์๋ฅผ ๋ค์ด, TV์ ๋ผ๋์ค๋ ๋งค์ถ๊ณผ ๊ด๋ จ์ด ์๋ค๋ ๊ฒ์ ์ด p-๊ฐ๋ค์ด ๋ํ๋ด์ง๋ง, ์ ๋ฌธ์ TV์ ๋ผ๋์ค๊ฐ ๊ณ ์ ๋ ์ํ์์ ๋งค์ถ๊ณผ ๊ด๋ จ์ด ์๋ค๋ ์ฆ๊ฑฐ๋ฅผ ์ ๊ณตํ๋ค.
๊ทธ๋ฌ๋ ์ด๋ฌํ ๊ฐ๋ณ p-๊ฐ๋ค์ ํตํด ๋ณ์๋ค๊ณผ ๋ฐ์ ๋ณ์ ์ฌ์ด์ ์ฐ๊ด์ฑ์ ๋ํด ๊ฒฐ๋ก ์ ๋ด๋ฆด ๋, ํนํ ์์ธก ๋ณ์ p์ ์๊ฐ ๋ง์ ๊ฒฝ์ฐ, ์ค๋ฅ๋ฅผ ๋ฒํ ๊ฐ๋ฅ์ฑ์ด ๋๋ค. ์ด์ ๋ ์ฐ์ฐํ ํ ๊ฐ ์ด์์ p-๊ฐ์ด 0.05 ์ดํ๋ก ๋์ฌ ํ๋ฅ ์ด ๋๊ธฐ ๋๋ฌธ์ด๋ค. ๊ทธ๋ฌ๋ F-ํต๊ณ๋์ ์์ธก ๋ณ์์ ์๋ฅผ ์กฐ์ ํจ์ผ๋ก์จ ์ด ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๋ค. ๋ฐ๋ผ์ ๊ท๋ฌด ๊ฐ์ค H0์ด ์ฐธ์ผ ๊ฒฝ์ฐ, ์์ธก ๋ณ์์ ์๋ ๊ด์ธก์น์ ์์ ์๊ด์์ด F-ํต๊ณ๋์ผ๋ก ์ธํด p-๊ฐ์ด 0.05 ์ดํ๋ก ๋์ฌ ํ๋ฅ ์ 5%์ ๋ถ๊ณผํ๋ค.
๋ฐ๋ผ์ ์ ์ฒด F-ํต๊ณ๋์ ์ดํด๋ณด๋ ๊ฒ์ด ์ค์ํ๋ค. ๊ฐ๋ณ ๋ณ์์ p-๊ฐ ์ค ํ๋๋ผ๋ ๋งค์ฐ ๋ฎ๋ค๋ฉด ์ ์ด๋ ํ๋์ ์์ธก ๋ณ์๊ฐ ๋ฐ์ ๋ณ์์ ๊ด๋ จ์ด ์๋ค๊ณ ๋ณด์ผ ์ ์์ง๋ง, ์ด๋ ์ค๋ฅ์ ๊ฐ๋ฅ์ฑ์ ๋์ธ๋ค. F-ํต๊ณ๋์ ์ฌ์ฉํ๋ฉด ์ด๋ฌํ ์ค๋ฅ ๊ฐ๋ฅ์ฑ์ ์ค์ด๊ณ , ๋ชจ๋ธ ์ ์ฒด๊ฐ ๋ฐ์ ๋ณ์์ ์ ์๋ฏธํ ๊ด๊ณ๋ฅผ ๊ฐ์ง๊ณ ์๋์ง ๋ณด๋ค ์ ํํ๊ฒ ํ๊ฐํ ์ ์๋ค.
'AI > ๋จธ์ ๋ฌ๋' ์นดํ ๊ณ ๋ฆฌ์ ๋ค๋ฅธ ๊ธ
ch2_learning2 ํธํฅ-๋ถ์ฐ Bias-variance Trade-off (0) | 2024.04.16 |
---|---|
Comparative Analysis of Classification Models: Logistic Regression, Naive Bayes, LDA, and QDA (0) | 2024.04.03 |
ch2_learning2 (Balancing Flexibility to Optimize Model Accuracy) ๋ชจ๋ธ์ ์ ํ๋ ํ๊ฐ (0) | 2024.03.13 |
ch2_learning (0) | 2024.03.12 |
ch1_overview (0) | 2024.03.06 |