๋ณธ๋ฌธ ๋ฐ”๋กœ๊ฐ€๊ธฐ

Statistics/Statistical Model

๋งค๊ฐœํšจ๊ณผ(Mediating effect)

โ–ก ๋งค๊ฐœํšจ๊ณผ(Mediating effect)

 

- ๋…๋ฆฝ๋ณ€์ˆ˜๊ฐ€ ์ข…์†๋ณ€์ˆ˜์— ์œ ์˜ํ•œ ์˜ํ–ฅ์„ ๋ฏธ์น  ๋•Œ, ๊ทธ ์‚ฌ์ด์— ๋งค๊ฐœ๋ณ€์ˆ˜๋ฅผ ํ†ตํ•ด ๊ทธ ์˜ํ–ฅ์ด ์ „๋‹ฌ๋˜๋Š” ๊ฒฝ์šฐ

- X(๋…๋ฆฝ๋ณ€์ˆ˜) → M(๋งค๊ฐœ๋ณ€์ˆ˜) → Y(์ข…์†๋ณ€์ˆ˜)

 

โ–ก ๋งค๊ฐœํšจ๊ณผ์˜ ์ข…๋ฅ˜

 

(1) ๋ถ€๋ถ„๋งค๊ฐœ ํšจ๊ณผ

 

 

(2) ์™„์ „๋งค๊ฐœ ํšจ๊ณผ

 

 

โ–ก ๋งค๊ฐœํšจ๊ณผ ๊ฒ€์ฆ๋ฐฉ๋ฒ•

 

- Baron & Kenny (1986) ๋ฐฉ๋ฒ•

 

     [1๋‹จ๊ณ„] X → M

     [2๋‹จ๊ณ„] X → Y

     [3๋‹จ๊ณ„] X + M → Y

 

(1) ๋ถ€๋ถ„๋งค๊ฐœ ํšจ๊ณผ ๊ฒ€์ฆ

 

     [1๋‹จ๊ณ„] M = a + b1*X → b1 ์œ ์˜

     [2๋‹จ๊ณ„] Y = a + phi*X → phi ์œ ์˜

     [3๋‹จ๊ณ„] Y = a + b2*X + b3*M → b3, b2 ์œ ์˜

 

โ–ท ์œ„์˜ 3๋‹จ๊ณ„์˜ ๊ฐ ๊ณ„์ˆ˜์˜ ์œ ์˜ ์กฐ๊ฑด์„ ๋งŒ์กฑํ•  ๊ฒฝ์šฐ, ๋ถ€๋ถ„๋งค๊ฐœํšจ๊ณผ๊ฐ€ ๋‚˜ํƒ€๋‚œ๋‹ค๊ณ  ๋ณผ ์ˆ˜ ์žˆ๋‹ค.

 

(2) ์™„์ „๋งค๊ฐœ ํšจ๊ณผ ๊ฒ€์ฆ

 

     [1๋‹จ๊ณ„] M = a + b1*X → b1 ์œ ์˜

     [2๋‹จ๊ณ„] Y = a + phi*X → phi ์œ ์˜

     [3๋‹จ๊ณ„] Y = a + b2*X + b3*M → b3๋งŒ ์œ ์˜

 

โ–ท ๋ถ€๋ถ„๋งค๊ฐœ ํšจ๊ณผ ๊ฒ€์ฆ๊ณผ์˜ ์ฐจ์ด์ ์€ 3๋‹จ๊ณ„์—์„œ b2๊ฐ€ ์œ ์˜ํ•˜์ง€ ์•Š๋Š”๋‹ค๋Š” ์ ์ด๋‹ค. ์ฆ‰, ์„ ํ˜•๋ชจ๋ธ์— ์„ค๋ช…๋ณ€์ˆ˜์™€ ๋งค๊ฐœ๋ณ€์ˆ˜๊ฐ€ ๊ฐ™์ด ๋“ค์–ด๊ฐ€ ์žˆ์„ ๋•Œ, ๋งค๊ฐœ๋ณ€์ˆ˜์˜ ๊ณ„์ˆ˜๋งŒ ์œ ์˜ํ•˜๋ฉด ์™„์ „๋งค๊ฐœ ํšจ๊ณผ๊ฐ€ ๋‚˜ํƒ€๋‚œ๋‹ค๊ณ  ๋ณผ ์ˆ˜ ์žˆ๋‹ค.

 

โ–ก ์ดํšจ๊ณผ

 

- ์ง์ ‘ํšจ๊ณผ์™€ ๊ฐ„์ ‘ํšจ๊ณผ์˜ ํ•ฉ

- ์ง์ ‘ํšจ๊ณผ: b2

- ๊ฐ„์ ‘ํšจ๊ณผ: b1*b3

 

     ์ดํšจ๊ณผ = ์ง์ ‘ํšจ๊ณผ + ๊ฐ„์ ‘ํšจ๊ณผ = b2 + b1*b3

 


Reference:

"11-3 ๋งค๊ฐœํšจ๊ณผ๋ž€ ๋ฌด์—‡์ธ๊ฐ€?," Sanpientia a Dei, www.youtube.com/watch?v=o7fF7uwi0s8&t=4s.

'Statistics > Statistical Model' ์นดํ…Œ๊ณ ๋ฆฌ์˜ ๋‹ค๋ฅธ ๊ธ€

ํ†ต๊ณ„ ๋ชจ๋ธ๋ง ๊ณผ์ •(Statistical modeling process)  (0) 2020.08.07
ํ†ต๊ณ„ ๋ชจ๋ธ(Statistical model)  (0) 2020.08.07