首页 | 学习中心 | 注册 登录
返回列表 发帖
RealEstate_CFA

Rank: 2

UID
223200 
帖子
247 
主题
139 
注册时间
2011-7-11 
最后登录
2017-8-12 
1# 跳转到 » 倒序看帖
打印
tT
发表于 2011-7-11 18:37 | 只看该作者

Null Hypothesis / Alternate Hypothesis

about, Robert, 2010
I am still confused about these.

Our poll says that candidate A is ahead of candidate B 52/48. What is the Null, what is the Alternate?

- Robert



Edited 1 time(s). Last edit at Friday, October 15, 2010 at 02:19AM by Robert A.
收藏 分享

disiz64

Rank: 2

UID
223312 
帖子
229 
主题
100 
注册时间
2011-7-11 
最后登录
2013-10-12 
2#
发表于 2011-7-11 18:37 | 只看该作者
Let me clarify my question.

ONE
-------------

I'm a medical researcher. I want to prove that a new drug is effective for treating a certain disease.

In setting up an hypothesis, there are two possible outcomes: The drug is effective in treating the disease, or the drug is not effective in treating the disease.

Which outcome is my null and which outcome is my alternate?

TWO
--------------

For years, it has been widely accepted that salt intake has no effect on weight. I believe this is incorrect and want to prove otherwise.

Which is the null hypothesis, that salt intake has no effect on weight, or that it does have an effect on weight?

Thank you,
Robert

- Robert

TOP

ramzes

Rank: 2

UID
223295 
帖子
246 
主题
96 
注册时间
2011-7-11 
最后登录
2013-8-21 
3#
发表于 2011-7-11 18:37 | 只看该作者
So, it's the alternative that I hope to prove true?

That is, Ha is true if the drug is effective, or that salt intake is related to weight gain?

- Robert

TOP

mnieman

Rank: 2

UID
223250 
帖子
258 
主题
116 
注册时间
2011-7-11 
最后登录
2013-10-9 
4#
发表于 2011-7-11 18:37 | 只看该作者
Another attempt to understand this....

You're prosecuting a case against a criminal defendant. You're trying to prove he's guilty of a crime. Are you trying to prove Ha (the alternate)?

- Robert

TOP

agulani

Rank: 2

UID
223253 
帖子
263 
主题
121 
注册时间
2011-7-11 
最后登录
2016-4-19 
5#
发表于 2011-7-11 18:37 | 只看该作者
Yes.

Let's say you want to prove the drug is effective. You would go:

H0=The drug isn't effective
H1=The drug is effective

So when you are able to reject H0 you can assure with a XX% confidence that the alternative (H1) is true, i.e. the drug is effective. If you are not able to reject H0 you just can't say anything about whether the drug it's effective or not.

The key is to know that you can prove H0 wrong, but never prove it true. Therefore, the statement you want to prove true must be H1.

In short:
-H1 is always the statement you hope to prove true.
-You can reject H0 => H1 is true // You can't reject H0 => you know nothing.

Just remember that and you'll be fine.

TOP

ishfaque

Rank: 2

UID
223317 
帖子
277 
主题
141 
注册时间
2011-7-11 
最后登录
2016-4-19 
6#
发表于 2011-7-11 18:37 | 只看该作者
Thanks. That clears a lot of confusion on the matter.

I'm a L2 candidate and I completely misunderstood this whole topic in L1. Ironically, I passed quant >70, probably by luck.

Now I have the time to learn it properly -- it comes up in L2.

- Robert

TOP

shootingstar

Rank: 2

UID
223278 
帖子
283 
主题
138 
注册时间
2011-7-11 
最后登录
2016-4-21 
7#
发表于 2011-7-11 18:37 | 只看该作者
Great stuff Fdez!

TOP

Maddin

Rank: 2

UID
223256 
帖子
209 
主题
93 
注册时间
2011-7-11 
最后登录
2016-4-18 
8#
发表于 2011-7-11 18:37 | 只看该作者
"So when you are able to reject H0 you can assure with a XX% confidence that the alternative (H1) is true, i.e. the drug is effective"

Just to muck things up a bit - When you reject H0, you are really saying that if H0 is true then you have drawn a sample so contrary to H0 that it would only happen 100*alpha % of the time. If H1 is the "omnibus" alternative (i.e., the one that includes everything that's not in H0) then either a rare event has happened or H1 is true. At some point (usu alpha = 0.05) we conclude that H0 is false because events rarer than 5% just aren't morally likely or something (We use 0.05 because RA Fischer one time said that 0.05 is no less arbitrary than any other number so we might as well use it).

The reason you can't exactly say that there is some probability that H1 is true is that H1 isn't a random event so it doesn't have probabilities associated with it.

Unless you're a Bayesian. And then you can say that stuff.

TOP

返回列表