Hypothesis Test: H0 & Ha - concepts

1. Hypothesis Testing? → Null Hypothesis(H0) 🙆‍♂️

1▶ Create a Hypothesis (without stating H0)

 

▷ (if data gives us strong evidence that the hypothesis is wrong) we can reject the Hypothesis

▷ (but when we have data that is similar to the hypothesis but not exactly the same) fail to reject the Hypothesis (best we can do)

(because it's unclear if the hypothesis should be based on many different results)

→ we need preliminary data in order to make a statement that we can test in follow up experiments

 

2▶ Null Hypothesis(귀무가설) - that there is no difference between things

 

ex) Null Hypothesis = there is no difference in recovery times btw two Drugs

▷ we simply see if the data convinces us to reject the Hypothesis that there is no difference btw two Drugs

▷ (if we tested the drugs on a lot of people & little random things would not change the results very much)

▷ we could confidently reject the Null Hypothesis (there is no difference btw two Drugs)

→ does not require preliminary data because the only value that represents no difference is 0

 

so.....

≫ rather than get stressed out over a large number of possible hypotheses... (1▶ 비추)

≫ use H0 to determine if there is a difference (2▶ 추천 - 즉 귀무가설을 사용해라~!)

 

cf) if you fail rejecting H0, (in ML world) it means that using two averages that you have overfit the data

2. Alternative Hypothesis (Ha)

※ statistical test

 

※ we need a STATISTICAL TEST to make a decision about whether or not to REJECT or NOT TO REJECT H0

 

※ a Statistical Test needs 3 things

→ data

→ a Null(or Primary Hypothesis) (i.e. need sth to reject or fail to reject)

→ Alternative Hypothesis (Ha) (simply the oppositve of H0)

ex) three drugs are not different from each other (H0)

- Ha would be... 'two drugs are the same & the remaining drug is doing its own thing'

- or.. Ha would be ... 'three drugs are all different from each other'

 

so.....

≫ depending on which one we use in the STATISTICAL TEST, we can end up making a different decision about H0

≫ so it is important to clearly state which Ha we want to use

(regardless of which Ha we chose, we only REJECT or FAIL to REJECT H0 or Primary Hypothesis)

---

(↓H0 & Ha를 이용한 하단 가설검정단계 기억하기↓)

 

1≫ if we tested the H0

2≫ ... using Ha, 

3≫ and we REJECT H0

{≫ (might say) we REJECTED H0 (in favor of this Ha)}

4≫ (BUT!) we would still not say we accept Ha (because other Ha MIGHT BE BETTER) (only if there are more than 2 optional Ha's)

(there are too many possibilities to test to know if we have accepted the correct one)

5≫ that is why we ONLY REJECT or FAIL TO REJECT H0 or Primary Hypothesis

---

* 출처1) https://www.youtube.com/watch?v=0oc49DyA3hU 

* 출처2) https://www.youtube.com/watch?v=5koKb5B_YWo  

* 썸네일출처) https://stock.adobe.com/sk/search?k=hypothesis%20icon