iSixSigma

Question About Null Hypothesis and Significance

Six Sigma – iSixSigma Forums General Forums General Question About Null Hypothesis and Significance

This topic contains 17 replies, has 8 voices, and was last updated by  Karthik Dharmalingam 5 days, 7 hours ago.

Viewing 18 posts - 1 through 18 (of 18 total)
  • Author
    Posts
  • #240565

    BayanKamal
    Participant

    Does the null hypothesis always assume there is no significance different on our data, even when we see/think that our data identical/same?

    0
    #240568

    Darth
    Participant

    The definition of “null” is having or associated with the value zero.  Therefore all null hypotheses will be in some form indicating an absence of something.  In the two sample t test for example, the null is that the difference in the population means of the two samples is zero or no difference.  In regression, the null hypothesis is that there is no regression.  Or that the slope is zero or the Y intercept is zero.

    Not sure what your question is: Does the null hypothesis always assume there is no significance different on our data, even when we see/think that our data identical/same?

    Your statement “assume there is no significance different on our data, even when we see/think that our data identical/same?”  doesn’t make sense.  You are saying the same thing.  There is however, a difference between statistical significance and mathematical difference.  The null refers to the population parameters while the sample values refer to the sample statistics.  We don’t care what the samples look like but what the data can tell us about the populations they came from.

    1
    #240577

    Mike Carnell
    Participant

    @bayankamal i agree with Darth’s comments.  Just a side not. If you go out and collect data on two separate occasions and end up with exactly the same values there is a real high probability there is something else going on. Something like a measurement system issue or something like that. Even if you end up with exactly the same mean and standard deviation you need to start kicking over rocks because it is a low probability.

    Just my opinion.

    1
    #240585

    Chris Seider
    Participant

    yes, just keep it simple…. hear null hypothesis and you should hear “equal” and the hypothesis is the statistical measure is equal to something else.

    even for a regression…the coefficients are equal to zero for the null hypothesis which means no regression exists

     

    1
    #240590

    Paul Chen
    Participant

    All great answers.  Here is another thought:

    You could also think about the Null Hypothesis as stating the status quo: that is, an assumption of there being no difference, and you have to come up with evidence to either reject this hypothesis or fail to reject it.

    The classic analogy is the criminal trial in which an assumption of innocence (i.e., a Null Hypothesis of not guilty) is made and it is up to the prosecution to present evidence that the jury must either fail to reject that Null Hypothesis (i.e., conclude the defendant is indeed not guilty), or reject the Null Hypothesis (i.e., and conclude that the defendant is indeed guilty).

    In the statistical test, the Null Hypothesis states there is no difference, and the statistical evidence (i.e., data) will show whether there is a difference or not, resulting in the statistical decision to reject or fail to reject that Null Hypothesis.

    0
    #240600

    Mike Carnell
    Participant

    @bayankamal I don’t know if this will help but iSixSigma ran this article earlier this year. Check the chart.

    https://www.isixsigma.com/community/blogs/the-history-of-the-hypothesis-testing-flow-chart/

     

    0
    #240622

    BayanKamal
    Participant

    Hi thankyou for the answer,

    Actually my question is more basic than your explanation, what i mean is if we claim that two part is same and we want to prove that this two part is actually different, what the Null Hypothesis should be?

    What im trying to ask is : does the null hypothesis always take the question that as null hypothesis in my case, the Null hypthesis : part A is not same with part B.

    • This reply was modified 1 month ago by  BayanKamal.
    0
    #240632

    Mike Carnell
    Participant

    @bayankamal The null hypothesis is always one equals the other.

     

    0
    #240643

    Darth
    Participant

    @bayankamal

    Let’s make sure that you are truly understanding what you are doing…..  The null hypothesis would be that the population mean that Sample A came from is the same as the population mean that Sample B came from.  You are NOT testing whether the means of the two samples are different.  The alternate hypothesis can be that the population mean of Sample A is different than the population mean that Sample B came from.  That is a two sided test.  You can also write the alternate hypothesis as the population mean of Part A is larger than the population mean of Part B or that the population mean of Part A is smaller than the population mean of Part B.  Those are one sided tests.  Finally you can also test alternate hypotheses such that the population means are different by some fixed number.

    1
    #240645

    BayanKamal
    Participant

    Hi @Darth & @mike-cernell

    Sorry if my question confuses both of you. I just start learning about this topic.

    Can I said that Null Hypothesis always equal to claim value? So if our claim is equal so our H0 is equal value, or if our claim is greater than our H0 will stated greater?

     

    0
    #240646

    Mike Carnell
    Participant

    @bayankamal Let me try an example. Let’s say I have two types of raw material, material A and material B. I want to test then to see if they process the same. H0 is Material A = Material B. It is always equal.

    0
    #240684

    Chuck White
    Participant

    @bayankamal, it sounds like you have a basic (but common) misunderstanding of how hypothesis testing works. A hypothesis test is similar to a proof by contradiction. In order to prove something logically or mathematically, a common approach is to assume the opposite and then demonstrate that this assumption results in a contradiction.

    In a hypothesis test, you basically do the same thing. If you want to prove that two samples came from different populations, you start with the assumption that they came from the same population — that is the null hypothesis. Then you analyze the data to see if it contradicts the assumption — if it does, you reject the null hypothesis and conclude the samples came from different populations.

    If your initial assumption was that the samples came from different populations, there wouldn’t be anything to contradict, so you couldn’t really prove anything.

    1
    #240695

    Chris Seider
    Participant

    @bayankamal fyi…your idea/question is NOT rarely asked.   Follow the advice of all above, including my previous post LOL, and you’ll proceed well.

    0
    #240696

    BayanKamal
    Participant

    @cseider yeah, I think i still need a long way to go,LOL. But thanks for all the answer, I learn six sigma from my company but they did not explain  detail of every part so there are a lot of things that I need to learn by myself to get comprehensive knowledge of six sigma.</p>

    • This reply was modified 3 weeks, 4 days ago by  BayanKamal.
    0
    #240702

    Mike Carnell
    Participant

    @bayankamal Welcome to the world of business buying cheap training from people who know just enough to get past HR. Then people wonder why deployments fail. No a lot of people have your energy to try to figure out what actually needs to be done.

    If you get good at hypothesis testing it will solve a lot of problems for you and the process people will love you. You can set this up and run it without,generally, disturbing the production process.

    0
    #240716

    Hajo Schmidt
    Participant

    Einstein once  said about assumption of nature laws:

    – In most cases the nature says: „You are wrong „ (we assume the alternative hypothesis)

    – In some cases the nature says: „Maybe you are right“ (we assume the null hypothesis)

    – But the nature never says: „Yes, you are right“ (two things are equal)

    By principle you can only state that something is not equal but you can never state that something is equal unless you know the entire population.

    • This reply was modified 3 weeks, 4 days ago by  Hajo Schmidt.
    0
    #240740

    Mike Carnell
    Participant

    @hajo I understand your comment. Basically it is the “fail to reject” comment. I have gone through this kind of discussion with my attorneys on several occasions. Regardless of what you say if you fail to reject the null hypothesis you behavior will be to treat them as if you had accepted the null hypothesis. You either treat them as if they are equal or like they are not.

    0
    #241124

    Karthik Dharmalingam
    Participant

    my 2 cents:

    Null is always “=”

    Alternate is “>” “<” “not equal” (unable to find a symbol :) )

    0
Viewing 18 posts - 1 through 18 (of 18 total)

You must be logged in to reply to this topic.