iSixSigma

Guidlines for critical P values

Six Sigma – iSixSigma Forums General Forums General Guidlines for critical P values

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #53822

    Watt
    Participant

    Hi All-

    I am looking for a document/book/URL that has a good, easy to understand, description of what critical P values are typically used in various industries or situations.

    I am not looking for a table to make my decisions for me, but rather a tool to help facilitate a discussion. I have been poking around and have not yet found something.

    Thanks!

    0
    #191559

    Robert Butler
    Participant

    There isn’t one and there isn’t a typical P-value for any industry or situation. There are defaults in both the literature and in software packages that exist due to a combination of convenience, tradition, and a sense of generalized acceptable risk but there is nothing industry or situation specific about them. What is true is that this inertia is such that if your choice of a critical P-value differs from one of the acceptable defaults you will usually be asked/required to explain/justify your choice.

    The Cartoon Guide to Statistics – Gonick and Smith pp.142 has this to say about defaults: ” In scientific work, a fixed alpha level of .05 or .01 is often used. These fixed levels are a holdover artifact from the pre-computer era, when we had to refer to tables, which were printed only for selected critical values. Still, many scientific journals continue to publish results only when the P-value is < .05."

    As to the why’s and wherefore’s concerning the old tabulation focus on .05 and .01 there are, I’m sure, any number of articles that might offer thoughts on this subject. If this is of interest to you I’d recommend pp.139-143 of An Introduction to Medical Statistics 3rd Edition – Bland as a starting point.

    0
    #191560

    Amour
    Participant

    The following URL offers an interesting discussion regarding critial values, specifically some historical perspective and various opinions on where to set you P-value.

    http://www.jerrydallal.com/LHSP/p05.htm

    0
    #191562

    MBBinWI
    Participant

    While there are “rules of thumb” the only good answer if that you need to determine the risk level for making an incorrect decision that you are willing to live with. Once made, stick with it. If you had been willing to accept a higher level of risk, then you should have specified that level of P value in the first place.

    Remember, there are no absolutes in statistics, so you are seeking a “how often could this happen entirely randomly, yet I still claim guilty.” If you must not make a mistake, then the P value must be correspondingly low (0.01 or even lower). If you’re willing to accept more chances that you’ll make a mistake, then set your P value higher (0.05 or 0.10 or even higher). The decision as to how much risk in “crying wolf” is up to you and the consequences of making the wrong call.

    0
    #191587

    Caco
    Participant

    What you described is the point of P-value.
    Thanks

    0
    #191588

    MBBinWI
    Participant

    Forest wrote:

    What you described is the point of P-value.
    Thanks

    So, did it help?

    0
    #191600

    Mikel
    Member

    Stats means never having to say you are certain.

    0
    #191603

    MBBinWI
    Participant

    Stan wrote:

    Stats means never having to say you are certain.

    And being an MBB means you can always answer “hmmm, it depends” LOL!

    0
Viewing 8 posts - 1 through 8 (of 8 total)

You must be logged in to reply to this topic.