# Can someone help me explain P-Value with a simple word please?

#### Berger

##### Registered
I tired to search on " What is P-value" and try to understand the meaning and how to interpret it, but I guess I am too dump to understand the term they use, "null hypothesis " , statistically highly significant, reject the null hypothesis etc.,
I still don't know if the P-value is less than 0.05 is consider good or bad, How do I know if the P-value I have is good?
For example below
on Example #1 : CPK is good but P-value is 0.907 what this mean? Can someone explain me ( please consider I am a dump person) please?

Example #2 : CPK is also good and P-Value < 0.005: How we can say about this? Why customer doesn't like to see if the P-value <0.005?

Sorry, if I asked non-sense or dump questions, but I admit that I am too dump to try to understand this by searching and read through google. I am apologies in advance. This is the first job I have to work with this.

Example #1 Example#2 #### Miner

##### Forum Moderator
Staff member
The p-value is a test on whether the data are normally distributed. If the p-value is greater than 0.05, you can assume that the data ARE normally distributed and your capability metrics (i.e., cp/Cpk...) are valid. If the p-value is less than 0.05, then you can assume that the data are NOT normally distributed and your capability metrics may not be valid.

The process shift in example 2 caused a mixture effect, which violates the normality test. This means the capability metrics are not valid.

#### blackholequasar

##### Involved In Discussions
Hey @Berger you're not stupid or dumb at all. Asking questions is a sign of intelligence! The folks here are a great and helpful resource. You'll learn! It all might seem intimidating, but you'll become familiar with a lot of these phrases and requirements. And someday, you'll be able to help another new person.

#### John Predmore

Trusted Information Resource
I had a stats professor who explained P-value simply, "if you reject the null hypothesis, the likelihood you would be wrong".

Most statistical tests are formulated so that the null hypothesis is the outcome you don't want, and there is a reason for that construction. The Anderson-Darling normality test (AD-Value) is a common exception where the null hypothesis is the data agree with a normal distribution, the outcome you likely want. Because the premise is reversed, the normality test reverses the common rule for other hypothesis testing. That is a lot of double negatives. It is easy to see why that might be confusing.

#### Miner

##### Forum Moderator
Staff member
I had a stats professor who explained P-value simply, "if you reject the null hypothesis, the likelihood you would be wrong".

Most statistical tests are formulated so that the null hypothesis is the outcome you don't want, and there is a reason for that construction. The Anderson-Darling normality test (AD-Value) is a common exception where the null hypothesis is the data agree with a normal distribution, the outcome you likely want. Because the premise is reversed, the normality test reverses the common rule for other hypothesis testing. That is a lot of double negatives. It is easy to see why that might be confusing.
I prefer to describe the null hypothesis as the status quo, and the alternate hypothesis as a change to the status quo. This avoids the reversal that you described.

p.s. I was trying to avoid using the technical jargon and keep it as simple as possible since that was what confused the OP.

#### Berger

##### Registered
The p-value is a test on whether the data are normally distributed. If the p-value is greater than 0.05, you can assume that the data ARE normally distributed and your capability metrics (i.e., cp/Cpk...) are valid. If the p-value is less than 0.05, then you can assume that the data are NOT normally distributed and your capability metrics may not be valid.

The process shift in example 2 caused a mixture effect, which violates the normality test. This means the capability metrics are not valid.
Hi Miner,
Thank you so much. The way you explained makes me understand the concept in just 1 mins compared with months try to understand on google search. I can see the picture now. I always think the P value < 0.005 is the best (which is wrong). Big thanks for you and I really appreciated it.

#### Berger

##### Registered
Hey @Berger you're not stupid or dumb at all. Asking questions is a sign of intelligence! The folks here are a great and helpful resource. You'll learn! It all might seem intimidating, but you'll become familiar with a lot of these phrases and requirements. And someday, you'll be able to help another new person.
Thank you so much Blackholequasar,
Thank you for understanding and yes, I will help other when possible as I know how it feel when you needed one. Thank you for kindness sir #### Berger

##### Registered
I had a stats professor who explained P-value simply, "if you reject the null hypothesis, the likelihood you would be wrong".

Most statistical tests are formulated so that the null hypothesis is the outcome you don't want, and there is a reason for that construction. The Anderson-Darling normality test (AD-Value) is a common exception where the null hypothesis is the data agree with a normal distribution, the outcome you likely want. Because the premise is reversed, the normality test reverses the common rule for other hypothesis testing. That is a lot of double negatives. It is easy to see why that might be confusing.
Thank you John, Yes, it's so confuse to me. I even asked my co-worker who seems to be expert in our company to explain me but he always use the term exactly on google where I don't understand. . Post on here even gives me a lot more information and understanding. Thanks sir #### Tidge

Trusted Information Resource