Statistics/Concepts(+codes)

Types of Errors in Hypothesis Testing

metamong 2022. 4. 27.

๐Ÿ‘„ ์ด๋ฒˆ ํฌ์ŠคํŒ…์—์„œ๋Š” ์šฐ๋ฆฌ๊ฐ€ ๋ฌด์กฐ๊ฑด ๋งˆ์ฃผํ•˜๊ฒŒ ๋˜๋Š” ๋‘ ๊ฐ€์ง€ ERROR! ํƒ€์ž…์— ๋Œ€ํ•ด์„œ ์•Œ์•„๋ณด๋ ค ํ•œ๋‹ค! ๐Ÿค

๐Ÿ‘„ ์ „์ฒด ๋ชจ์ง‘๋‹จ population์— ๋Œ€ํ•œ ์ถฉ๋ถ„ํ•œ ์ •๋ณด๊ฐ€ ์ œ๊ณต๋˜์ง€ ์•Š์•˜๊ธฐ ๋•Œ๋ฌธ์—! ์šฐ๋ฆฌ๋Š” sample์„ ์ด์šฉํ•˜์—ฌ ์ „์ฒด population์— ๋Œ€ํ•œ ์ถ”๋ก ์„ ํ•˜๋Š” ๊ณผ์ •์„ ๊ฑฐ์นœ๋‹ค. ์ด๋Ÿฐ ์ถ”๋ก  ๊ณผ์ •์„ ์ผ์ข…์˜ hypothesis testing์ด๋ผ ํ•˜๋ฉฐ, ์ด testing ๊ณผ์ •์—์„œ ์šฐ๋ฆฌ๋Š” ๋‹น์—ฐํžˆ sample๋งŒ์„ ๊ฐ€์ง€๊ณ  ์ถ”๋ก ํ•˜๊ธฐ์— ์—ฌ๋Ÿฌ error์— ๋งˆ์ฃผ์น˜๊ฒŒ ๋จ

 

 

Hypothesis Test: H0 & Ha - concepts

1. Hypothesis Testing? → Null Hypothesis(H0) ๐Ÿ™†‍โ™‚๏ธ 1โ–ถ Create a Hypothesis (without stating H0) โ–ท (if data gives us strong evidence that the hypothesis is wrong) we can reject the Hypothes..

sh-avid-learner.tistory.com

 

๊ฐœ๋…>

๐Ÿค™ ์ œ์ผ ์ค‘์š”ํ•œ ๊ฑด H0! H0๋ฅผ rejectํ•จ์œผ๋กœ์จ ํ•ด๋‹น ๊ฐ€์„ค์˜ ์‹ ๋น™์„ฑ effect๊ฐ€ ์žˆ๋‹ค๊ณ  ๊ฒฐ๋ก ์„ ๋‚ด๋ฆฌ๊ฒŒ ๋˜๋Š” ๊ณผ์ •์ด ํ•„์š”ํ•œ๋ฐ, ์ด hypothesis test์—์„œ ํฌ๊ฒŒ ๋‘ ๊ฐ€์ง€์˜ error๋ฅผ ๋งˆ์ฃผํ•˜๊ฒŒ ๋จ!

 

 

๐Ÿค™ Type1 error (FP; False Positive)

โ–ถ hypothesis testing ๊ฒฐ๊ณผ ์šฐ๋ฆฌ๊ฐ€ ์„ธ์šด H0๊ฐ€ ๋ฐ”๋ผ๋˜ ๋Œ€๋กœ reject์ด ๋˜์—ˆ๋‹ค. ์ฆ‰ Ha๋ฅผ ์„ ํƒํ•  ์ฐจ๋ก€. ํ•˜์ง€๋งŒ ์‹ค์ œ ์ƒํ™ฉ์„ ๋ณด๋‹ˆ H0 ๋‚ด์šฉ์ด reject๋˜์–ด์„œ๋Š” ์•ˆ๋˜๊ณ , H0๊ฐ€ ์„ฑ๋ฆฝ๋˜์–ด์•ผ ๋งž๋‹ค๊ณ  ๋‚˜์˜จ๋‹ค. ์ด๋ ‡๊ฒŒ ๋ฐœ์ƒํ•˜๋Š” ์—๋Ÿฌ๋ฅผ ์šฐ๋ฆฌ๋Š” Type 1 error๋ผ ๋ถ€๋ฅด๊ณ  H0์„ rejectํ•˜์ง€ ๋ง์•„์•ผ ํ•œ๋‹ค!

โ–ถ FP๋ผ๊ณ ๋„ ๋ถ€๋ฅธ๋‹ค. Positiveํ•˜๊ฒŒ Ha๋ฅผ ์„ ํƒํ•œ ์ƒํ™ฉ์ด False๋กœ ๋‚˜์™”์Œ ๐Ÿฅบ

 

๐Ÿค™ Type2 error (FN; False Negative)

โ–ถ hypothesis testing ๊ฒฐ๊ณผH0์„ rejectํ•ด์„œ๋Š” ์•ˆ๋˜๊ณ  H0์„ ๋ฐ›์•„๋“ค์—ฌ์•ผ ํ•˜๋Š” ์ƒํ™ฉ์ด ๋‚˜์™”๋‹ค. ํ•˜์ง€๋งŒ ์‹ค์ œ ์ƒํ™ฉ์—์„œ๋Š” H0์€ ๊ฑฐ์ง“์œผ๋กœ ํŒ๋ช…. ์šฐ๋ฆฌ๋Š” ์ด๋Ÿด ๋•Œ Type 2 error๊ฐ€ ๋ฐœ์ƒํ–ˆ๋‹ค ๋ถ€๋ฅด๊ณ  H0์— ๋ฐ˜ํ•˜๋Š” ์‚ฌ์‹ค์„ ์ฑ„ํƒํ•ด์•ผ ํ•œ๋‹ค!

โ–ถ FN์ด๋ผ๊ณ ๋„ ๋ถ€๋ฅธ๋‹ค. Negativeํ•˜๊ฒŒ H0์„ ์„ ํƒํ•œ ์ƒํ™ฉ์ด False๋กœ ํŒ๋ช…๋‚ฌ๊ธฐ ๋•Œ๋ฌธ! ๐Ÿฅบ

 

์˜ˆ์‹œ>

Q1) ๋ถˆ์ด ์—†๋Š”๋ฐ ํ™”์žฌ ๊ฒฝ๋ณด์Œ์ด ์šธ๋ ธ๋‹ค๋ฉด? FP 

Q2) ๋ถˆ์ด ์žˆ๋Š”๋ฐ ํ™”์ œ ๊ฒฝ๋ณด์Œ์ด ์šธ๋ฆฌ์ง€ ์•Š์•˜๋‹ค๋ฉด? FN

Type 1 error>

๐Ÿ‘‹ ํ†ต๊ณ„์  ๊ด€์ ์—์„œ ๋งˆ์ฃผํ•˜๋Š” Type 1 error ์ƒํ™ฉ์„ ์‚ดํŽด๋ณด์ž!

 

โ˜๏ธ ๊ณ„์‚ฐ ๊ฒฐ๊ณผ ๋‚˜์˜จ p-value๊ฐ€ significance level ์œ ์˜์„ฑ ๋ฒ”์œ„๋ณด๋‹ค ์ ๊ฒŒ ์‚ฐ์ถœ๋˜์—ˆ๋‹ค. → ์ ๊ฒŒ ์‚ฐ์ถœ๋˜์—ˆ์œผ๋ฏ€๋กœ ์šฐ๋ฆฌ๊ฐ€ ๊ฒ€์ •ํ•˜๋Š” ์˜๋„๋Œ€๋กœ H0์ด reject๋จ  H0์„ rejectํ•˜๋Š” ์ƒํ™ฉ์ด ์ƒ๊ฒจ hypothesis testing์ด ์˜๋ฏธ๊ฐ€ ์žˆ๋‹ค(statistically significant)๊ณ  ์ƒ๊ฐํ•  ์ˆ˜ ์žˆ๋‹ค.

ํ•˜์ง€๋งŒ ์ด ๋•Œ, sample error๋กœ ์ธํ•ด FP ๋˜๋Š” ์ƒํ™ฉ์ด ์ƒ๊ธธ ์ˆ˜ ์žˆ์Œ. sample์„ ํ†ตํ•ด์„œ ๊ฒ€์ •ํ•˜๋ ค๊ณ  ํ•˜๋Š” supposed effect๊ฐ€ ์‹ค์ œ population์—์„œ๋Š” ์กด์žฌํ•˜์ง€ ์•Š์„ ์ˆ˜ ์žˆ๊ธฐ ๋•Œ๋ฌธ

 

- p-value & significance level์— ๋”ฐ๋ฅธ ๊ฒฐ๊ณผ ํ•ด์„ ์ฐจ์ด -

 

'If the p value of your test is lower than the significance level, it means your results are statistically significant and consistent with the alternative hypothesis. If your p value is higher than the significance level, then your results are considered statistically non-significant.'

 

โœŒ๏ธ ์ด Type 1 error์˜ ๋นˆ๋„ the rate of occurence๋ฅผ ์•Œ ์ˆ˜ ์žˆ์„๊นŒ? error ๋นˆ๋„๋Š” ์œ ์˜์ˆ˜์ค€ α์™€ ๊ฐ™๋‹ค. 

 

'The significance level is an evidentiary standard that you set to determine whether your sample data are strong enough to reject the null hypothesis. Hypothesis tests define that standard using the probability of rejecting a null hypothesis that is actually true. You set this value based on your willingness to risk a false positive.'

 

ex) ์˜ˆ๋ฅผ ๋“ค๋ฉด ์œ ์˜์ˆ˜์ค€ α๊ฐ€ 0.05์ด๋ผ๋ฉด, FP๊ฐ€ ๋ฐœ์ƒํ•  ํ™•๋ฅ ์€ 5%์ด๋ฉฐ, ์ด๋Š” ๊ณง H0์ด reject๋˜์–ด์„œ๋Š” ์•ˆ๋˜๊ณ  H0์ด ์‹ค์ œ true๋กœ ํŒ๋ช…๋‚  ํ™•๋ฅ 

→ ์œ ์˜์ˆ˜์ค€ α ์ž์ฒด๊ฐ€ H0์„ rejectํ•˜๋Š” ์ง€์— ๋Œ€ํ•œ ๊ธฐ์ค€์„ ์„ธ์šด ๊ฒƒ์ด๊ธฐ์—, FP๋ผ๊ณ  ๋‚ด๋ฆด ์ˆ˜ ์žˆ๋Š” ๊ธฐ์ค€์ด๋ผ ํ•  ์ˆ˜ ์žˆ๊ณ , ์ด๋Š” ๊ณง FP์˜ ๋นˆ๋„๊ฐ€ ๋˜๋Š” ๊ฒƒ!

 

ex - ์‹ค์ œ clinical study

'In your clinical study, you compare the symptoms of patients who received the new drug intervention or a control treatment. Using a t test, you obtain a p value of .035. This p value is lower than your alpha of .05, so you โ‘ consider your results statistically significant and reject the null hypothesis. However, the p value means that โ‘กthere is a 3.5% chance of your results occurring if the null hypothesis is true. Therefore, there is still a risk of making a Type I error.'

Type 2 error>

๐Ÿ‘‹ Type1๊ณผ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ ํ†ต๊ณ„์  ๊ด€์ ์—์„œ Type2 ์ƒํ™ฉ์„ ํ™•์ธํ•ด๋ณด์ž

 

โ˜๏ธ ์ด์   ๊ณ„์‚ฐ ๊ฒฐ๊ณผ ๋‚˜์˜จ p-value๊ฐ€ significance level ์œ ์˜์„ฑ ๋ฒ”์œ„๋ณด๋‹ค ๋” ๋†’๊ฒŒ ์‚ฐ์ถœ๋˜์—ˆ๋‹ค. → ๋” ๋†’๊ฒŒ ์‚ฐ์ถœ์ด ๋˜์—ˆ๋‹ค๋Š” ๊ฑด, hypothesis testing์—์„œ H0์€ reject๋˜์ง€ ์•Š์Œ. H0 ๊ท€๋ฌด๊ฐ€์„ค์„ ๋ฐ›์•„๋“ค์ด๋Š” ๊ฒƒ์ด๋ฏ€๋กœ, ํ•ด๋‹น testing์˜ ํšจ๊ณผ effect๊ฐ€ ์—†๋‹ค๋Š” ๊ฒƒ์ด๊ธฐ์—  ํ•ด๋‹น sample์ด ์›ํ•˜๋Š” effect๋ฅผ ํŒ๋‹จํ•˜๊ธฐ ์œ„ํ•ด ์ถฉ๋ถ„ํ•˜์ง€ ์•Š๋‹ค(not having enough statistical power)๊ณ  ๊ฒฐ๋ก ์ด ๋‚˜์˜จ ๊ฒƒ์ด๋‹ค.

ํ•˜์ง€๋งŒ type 2 error๋กœ ์ธํ•ด ๋ณด์ด๊ณ ์ž ํ•˜๋Š” ๊ฒฐ๋ก ์ด ์œ ํšจํ•  ๊ฐ€๋Šฅ์„ฑ(H0์ด reject๋˜๋Š”, ํšจ๊ณผ๊ฐ€ ์žˆ๋‹ค๋Š”)์ด ์žˆ๋‹ค.

์ด ๊ฐ€๋Šฅ์„ฑ์„ β(error rate)๋ผ๊ณ  ํ•œ๋‹ค.

(** ์—ฌ๊ธฐ์„œ statistical power๋Š” ์‹ค์ œ H0์ด false์ธ ๊ฐ€์šด๋ฐ, ์‹ค์ œ๋กœ H0์„ rejectํ•  ํ™•๋ฅ ์ด๋‹ค)

 

โœ‹ ์ž ๊น! ์œ„์—์„œ์˜ statistical power๋Š” test์—์„œ ์šฐ๋ฆฌ๊ฐ€ ์•Œ๊ณ ์ž ํ•˜๋Š” ์‹ค์ œ ์˜ํ–ฅ์ด ์žˆ๋Š” ์ง€๋ฅผ ํŒ๋‹จํ•˜๋Š” ์ •๋„(testing์ด ์˜๋ฏธ๊ฐ€ ์žˆ๋Š” ์ง€)๋ฅผ ๋œปํ•จ. (์ฆ‰ H0์ด rejected๋˜์–ด ์‹ค์ œ ํšจ๊ณผ๊ฐ€ ์žˆ๋‹ค๊ณ  ๋ณด๋Š” test์˜ ๊ฒ€์ •ํšจ๊ณผ) - ์ฃผ๋กœ 80% power level ์ด์ƒ์ผ ๋•Œ ์‹ค์ œ ํšจ๊ณผ ์žˆ๋‹ค ํŒ๋‹จ

 

ex - ์‹ค์ œ clinical study(2) -

'When preparing your clinical study, you complete a power analysis and determine that with your sample size, you have an 80% chance of detecting an effect size of 20% or greater. An effect size of 20% means that the drug intervention reduces symptoms by 20% more than the control treatment. However, a Type II may occur if an effect that’s smaller than this size. A smaller effect size is unlikely to be detected in your study due to inadequate statistical power.'

 

โœŒ๏ธ Type1๊ณผ ๋‹ค๋ฅด๊ฒŒ ์—ฌ๋Ÿฌ ์š”์ธ๋“ค๋กœ ์ธํ•ด Type2๊ฐ€ ๋ฐœ์ƒํ•œ๋‹ค. ๋˜ํ•œ occurrence์ธ β๋ฅผ ์•Œ๊ธฐ๋„ ๋งค์šฐ ์–ด๋ ต๋‹ค. ์ง์ ‘ Ha์— ๊ด€ํ•œ study๋ฅผ ์ง„ํ–‰ํ•˜๊ณ  Ha distribution์„ ํ†ตํ•ด ์–ด๋ฆผ์žก์•„ estimate์„ ์•Œ ์ˆ˜ ์žˆ์„ ์ •๋„.

 

Q. ๊ทธ๋ ‡๋‹ค๋ฉด Statistical Power ๊ฒฐ์ • ์š”์ธ์€?

A1) size of the effect (effect size ์ปค์ง€๋ฉด power ์ปค์ง)

A2) measurement error (systematic & random errors) - error ๋ฐœ์ƒํ• ์ˆ˜๋ก power ๊ฐ์†Œ

A3) sample size (๋‹น์—ฐํžˆ sample size๊ฐ€ ์ปค์ง€๋ฉด power๋Š” ์ปค์ง)

A4) significance level (level๋„ ์ปค์ง€๋ฉด power๋„ ์ปค์ง)

 

→ low variability & larger effect size๋ผ๋Š” ์š”์ธ์œผ๋กœ Type2 ๋ฐœ์ƒ ๋น„์œจ์„ ์ค„์ผ ์ˆ˜ ์žˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์‹ค์ œ ๊ฒ€์ •์—์„œ๋Š” variability & effect size๋ฅผ ์ปจํŠธ๋กคํ•˜๊ธฐ์—๋Š” ์–ด๋ ค์šฐ๋ฏ€๋กœ sample size๋ฅผ ์ฆ๊ฐ€์‹œํ‚ค๊ฑฐ๋‚˜ significance level์„ ์˜ฌ๋ ค์„œ Type2 ๋ฐœ์ƒ ๋น„์œจ์„ ์ค„์ด๋ ค๊ณ  ํ•œ๋‹ค!

Errors in distributions

* H0 distribution>

H0์ด true๋ผ๊ณ  ๊ฐ€์ •ํ–ˆ์„ ๋•Œ, ์ƒˆ๋กœ์šด sample๋“ค์„ ๊ณ„์† ๋ฐ˜๋ณตํ•˜๋ฉด์„œ ๋‚˜์˜ฌ ์ˆ˜ ์žˆ๋Š” ๋ชจ๋“  ์ƒํ™ฉ๋“ค์ด ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋Š” ํ™•๋ฅ ๋“ค์˜ ๋ชจ์Œ์ง‘

→ α๋ผ๊ณ  ์“ฐ์—ฌ์ง„ ๋ถ€๋ถ„์€ critical region์ด๋ผ ๋ถ€๋ฅด๋ฉฐ ํ•ด๋‹น region์— ๊ฒฐ๊ณผ๊ฐ€ ํฌํ•จ๋˜๋ฉด, ๋‹น์—ฐํžˆ H0์€ rejected๋˜๋Š”, statistically significantํ•œ ๋ถ€๋ถ„์ด๋ผ๊ณ  ๋งํ•  ์ˆ˜ ์žˆ๋‹ค. ๊ทผ๋ฐ distribution ์ž์ฒด๊ฐ€ H0์ด trueํ•œ ์ƒํ™ฉ์ด๊ธฐ์—, FP๋กœ ํŒ๋‹จ๋˜๋Š” region์ด๋‹ค!

 

-- ์ตœ์ข… ์ •๋ฆฌ --

 

<Type 1 error>

 

→ hypothesis testing ์‹œํ–‰

→ p-value๊ฐ€ significance level๋ณด๋‹ค ์ž‘๊ฒŒ ์ธก์ •๋จ

→ H0์ด reject๋จ

→ ์˜๋„๋Œ€๋กœ testing ๊ฒ€์ • ํšจ๊ณผ๋ฅผ ๋ด„

→ ์šฐ๋ฆฌ๋Š” statistically significantํ•˜๋‹ค๊ณ  ๊ฒฐ๋ก ์„ ๋‚ด๋ ธ์Œ

→ ๊ทผ๋ฐ sample error๋กœ type 1 error ๋ฐœ์ƒ

→ ์‹ค์ œ๋กœ๋Š” H0์ด true๋ผ๊ณ  ํ•จ

→ ์•„, ๊ฒฐ๊ตญ ์šฐ๋ฆฌ๋Š” STATISTICALLY SIGNIFICANT(POSITIVE)๋ผ๊ณ  ๊ฒฐ๋ก ์„ ๋‚ด๋ฆฐ ๊ฑธ ์ž˜๋ชป(FALSE) ํŒ๋‹จํ–ˆ์Œ์„ ์‹œ์ธ;; ๐Ÿ˜”

F(alse)P(ositive)

 

๐Ÿ’ช FALSEํ•˜๊ฒŒ POSITIVE(STATICALLY SIGNIFICANT)๋กœ ํŒ๋‹จ

 

* Ha distribution>

→ ์—ญ์‹œ H0์™€ ๊ฐ™์€ ์›๋ฆฌ๋กœ Ha๊ฐ€ true๋ผ๊ณ  ๊ฐ€์ •ํ–ˆ์„ ๋•Œ์˜ ์ƒˆ๋กœ์šด sample๋“ค์„ ๋ฐ˜๋ณตํ•œ ๊ฒฐ๊ณผ ๋‚˜์˜ฌ ์ˆ˜ ์žˆ๋Š” ์ƒํ™ฉ ํ™•๋ฅ ๋“ค ๋ชจ์Œ์ง‘

 

-- ์ตœ์ข… ์ •๋ฆฌ --

 

<Type 2 error>

 

→ hypothesis testing ์‹œํ–‰

→ p-value๊ฐ€ significance level๋ณด๋‹ค ํฌ๊ฒŒ ์ธก์ •๋จ

→ H0์ด reject๋˜์ง€ ์•Š์Œ

→ ์šฐ๋ฆฌ๊ฐ€ ์ƒ๊ฐํ•˜๋˜๋Œ€๋กœ testing ํšจ๊ณผ๋Š” ์—†๋‹ค ํŒ๋‹จํ•จ

→ statistical power๊ฐ€ ์—†๋Š” test๊ตฌ๋‚˜.. ใ… ใ… 

→ ๊ทผ๋ฐ ์—ฌ๋Ÿฌ ์š”์ธ๋“ค๋กœ ์ธํ•ด type 2 error ๋ฐœ์ƒ

→ ์‹ค์ œ๋กœ๋Š” H0์ด reject๋˜์–ด์•ผ ํ•œ๋‹ค๊ณ  ํ•จ

→ ์•„, ์šฐ๋ฆฌ๋Š” statistical power๊ฐ€ ์—†๋‹ค๊ณ (NEGATIVE) ๊ฒฐ๋ก ์„ ๋‚ด๋ ธ๋Š”๋ฐ

→ ์‹ค์ œ๋Š” ์•„๋‹ˆ๊ตฌ๋‚˜(FALSE)์ž„์„ ์‹œ์ธ;;

 F(alse)N(egative)

 

๐Ÿ’ช FALSEํ•˜๊ฒŒ NEGATIVE(NOT HAVING STATISTICAL POWER)๋กœ ํŒ๋‹จ

 

 

type1๊ณผ type2 ๊ฐ„์˜ TRADE-OFFS

๐Ÿ‘ ๋‘ type error๋Š” ์„œ๋กœ ์˜ํ–ฅ์„ ๋ฐ›๋Š” trade-off ๊ด€๊ณ„์ด๋‹ค

 

โ‘  significance level(์œ„ alpha์˜ ๋ฉด์ )์„ ์ค„์ด๋ฉด? → testing์—์„œ p-value๊ฐ€ significance level๋ณด๋‹ค ์ ๊ฒŒ ์ธก์ •๋  ํ™•๋ฅ ์ด ๊ฐ์†Œํ•œ๋‹ค → H0์ด reject๋  ํ™•๋ฅ ์ด ๊ฐ์†Œํ•œ๋‹ค ์‹ค์ œ๋กœ H0์ด true์ธ ์ƒํ™ฉ์—์„œ H0์ด true๊ฐ€ ์•„๋‹ˆ๋ผ๊ณ  ํŒ๋‹จํ•  ํ™•๋ฅ (type 1 error)์ด ๊ฐ์†Œํ•œ๋‹ค type1 error ๋ถ€๋‹ด์ด ์ค„์–ด๋“ ๋‹ค / ๋ฐ˜๋Œ€๋กœ type 2 error ๋ถ€๋‹ด์ด ์ฆ๊ฐ€ํ•œ๋‹ค

โ‘ก test์˜ statistical power๋ฅผ ์ฆ๊ฐ€์‹œํ‚ค๋ฉด? → testing ํšจ๊ณผ๊ฐ€ ์ฆ๊ฐ€ํ•œ๋‹ค → ์ฆ‰, ๋˜‘๊ฐ™์€ ๋ง๋กœ H0์„ rejectํ•  ํ™•๋ฅ ์ด ์ฆ๊ฐ€ํ•œ๋‹ค ์‹ค์ œ Ha๊ฐ€ true์ธ ์ƒํƒœ(H0์ด true๊ฐ€ ์•„๋‹Œ ์ƒํƒœ)์—์„œ H0์„ ๊ณ ๋ฅผ ํ™•๋ฅ (type 2 error)์ด ๊ฐ์†Œํ•œ๋‹ค type2 error ๋ถ€๋‹ด์ด ์ค„์–ด๋“ ๋‹ค / type1 error ๋ถ€๋‹ด์ด ์ฆ๊ฐ€ํ•œ๋‹ค

 

โ‰ซ ์ฆ‰ ํ•œ ์ชฝ error ๋ถ€๋‹ด์„ ์ค„์ด๋ฉด ๋‹ค๋ฅธ error ํ™•๋ฅ ์ด ์ฆ๊ฐ€๋˜๋Š” ์„œ๋กœ trade-off ๊ด€๊ณ„์ž„์„ ์œ„ ๊ทธ๋ฆผ์„ ํ†ตํ•ด์„œ๋„ ์•Œ ์ˆ˜ ์žˆ์Œ!

 

- ์–ด๋ ค์šด ํ†ต๊ณ„ ๊ฐœ๋… ์ •๋ฆฌ ๋! ํœด ๐Ÿ˜‡-


* ์ธ๋„ฌ ์ถœ์ฒ˜) https://liwaiwai.com/2019/10/12/statistics-for-dummies-type-i-and-type-ii-errors/

* ๋‚ด์šฉ & ๊ทธ๋ฆผ ์ถœ์ฒ˜1) https://www.publichealthnotes.com/hypothesis-testing-and-difference-between-type-i-and-type-ii-error/type-1-and-2-errors/

* ์ถœ์ฒ˜2) https://statisticsbyjim.com/hypothesis-testing/types-errors-hypothesis-testing/

* ์ถœ์ฒ˜3) https://www.scribbr.com/statistics/type-i-and-type-ii-errors/

๋Œ“๊ธ€