# P – Value and its interpretation

Six Sigma – iSixSigma Forums Old Forums General P – Value and its interpretation

Viewing 11 posts - 1 through 11 (of 11 total)
• Author
Posts
• #40479

Tanmoy
Member

What does P value in Regression signify . How to interpret it ? I am confused about P- value as it is used in hypothesis testing and  in other tools such as Kruskal – Wallis Test or Regression .

0
#125741

Jaybee
Participant

Hi Tanmoy,
I am feeling in a generous mood today (don’t know why as the day at work has been garbage) anyhoo, post your e-mail address on here & i will send you an example of a full regression model – although its manufacturing based, it doesn’t really matter however.
It will be tomorrow sometime (if i haven’t emigrated by then!)
Jaybee

0
#125773

ripper
Member

The p-value tells you the same thing no matter where you encounter it, be it in a t-test, ANOVA, regression, experiment, etc.  That’s the beauty of it…It you have a p-value, you have a hypothesis.  If you have a hypothesis, you have some statement you are trying to dissprove.  You can’t determine the direction or magnitude of some value using the p-value, only if something is or isnt.  The key is to always know what that something is (null) and what it isnt (alternative).  The p-value is then measured against the level of significance you selected to determine if you are rejecting or accepting the null – “If the p-value is low (p-value < alpha value), the null must go. (practically, not statistically speaking).
Hence, you are determining, “This thing is what my null says it is, or it isn’t”. I find it highly useful to always write your null and alternative hypothesis when you are first starting out.
I am a fetal practicianer, so confirm this explanation.  Research under “hypothesis testing, alpha, type I or II, probability statistics, etc”.  All relate directly to your question. Good luck.

0
#125778

HF Chris
Participant

Having a P-Value > alpha (dependent whether liberal or conservative) for the null only indicates that the null was not accepted…not rejected.  Unless you have the entire population you are dealing with a sample (how representative is your sample?). Although semantics in nature, a lot of good projects have not been shared because the null was rejected.
Chris

0
#125818

CSSBB
Participant

Chris,
You would never make a statement “Null is not accepted”. You either reject a null or fail to reject it.
Just my opinion.

0
#125822

cgvallee
Participant

When the stats were ran on the space shuttle o-rings, I believe they rejected the null. Look what happened when the whole sample was not present.
Chris

0
#125823

ripper
Member

Certainly true…but the question was scripted in such a way as to indicate someone with a very basic understanding of the concept, thus insisting on proper semantics at this point in their learning curve appeared counterproductive.  Hence the reason I indicated my explanation was more practical than statistical.

0
#125824

Robert Butler
Participant

While this is off topic I wouldn’t use the “analysis” of the shuttle o-ring data for any kind of discussion about sample analysis because there weren’t any stats worthy of the name run by anyone connected with that nightmare.  The people who were evaluating the data had no real understanding of what data analysis meant.  The real tragedy was that they had “an inability to assess the link between cool temperature and O-ring damage on earlier flights. Such a pre-launch analysis would have revealed that this flight was at considerable risk. On the day before the launch of Challenger, the rocket engineers and managers needed a quick, smart analysis  of evidence about the threat of cold to the O-rings, as well as an effective presentation of evidence in order to convince NASA offocials not to launch.” (pp.40 Visual Explanations – Tufte)
If you want to see what the engineers sent and read about how it was presented and also see how the exact same data could have been presented in a manner guaranteed to draw attention to the risks I know of no better discussion of the issue than pp. 39-53 of the book mentioned above.

0
#125827

HF Chris
Participant

Robert,
You could not have posted it any better. The statement ” The people who were evaluating the data had no real understanding of what data analysis meant, ” was exactly what I was driving towards. At this point in the game, semantics can make a difference when one understands truely what it is and is not.
Chris

0
#125833

BTDT
Participant

Robert:A great book. I saw a copy in the window of a used bookstore about a year back for \$35. I went 1/2 block down the street, turned around to get it, to find it had just been sold.Non carpe diem.BTDT

0
#125965

Tanmoy
Member

A new message by Jaybee was posted in the Discussion Forum.

Hi Tanmoy,
Thanks . Its [email protected]

I am feeling in a generous mood today (don’t know why as the day at work has been garbage) anyhoo, post your e-mail address on here & i will send you an example of a full regression model – although its manufacturing based, it doesn’t really matter however.
It will be tomorrow sometime (if i haven’t emigrated by then!)
Jaybee

0
Viewing 11 posts - 1 through 11 (of 11 total)

The forum ‘General’ is closed to new topics and replies.