iSixSigma

P value of .05 instead of a P value of .01

Six Sigma – iSixSigma Forums Old Forums General P value of .05 instead of a P value of .01

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #39956

    Niki
    Participant

          Why do we generally use a P value of .05 or a confidence interval of 95% instead of a P value of .01 or a confidence interval of 99%?

    0
    #122835

    BTDT
    Participant

    niki:R.A. Fisher (1890-1962) developed many of the statistical techniques that are common in Analyze. When he was describing the results of an agricultural test, he stated,“either there is something in the treatment, or a coincidence has occurred such as does not occur in 1 in 20 trials”The 1 in 20 rule convinces most people and has been generally accepted ever since. There are some groups that don’t go with the 0.05 (95%CI). The social sciences will frequently publish results using 0.10 (90% CI). The nuclear industry tends to go with 0.01 (99% CI), and I’m sure we are all glad about that.(Journal of the Ministry of Agriculture of Great Britain, 1926)BTDT

    0
    #122840

    anurag kabthiyal
    Participant

    Dear Niki,
    Statistics is all about prediction (inferential statistic)…..let say u have to catch a bus which is at other side of the road , for that u analyze the situation n then make a risk assessment , how much ll be mine speed so that I catch the bus and what are the chance at that speed so i can catch the bus……u make an estimation of speed say if i run at a speed of 80KPH to 100kph …99% chance i ll catch it……if  suppose i ask u to run @ speed 50kph to 70kph….what ll be urs chance of catching the bus ……?…u ll say 90% or 85%
    so in statistic we say thing based on the how confident we are …..How confident we want to be in that situation…..or how much risk associated ..or covered …..if risk is less i can choose the p= 0.1(90%) in case of social sciences survey……if it is critical we generally (based on experience ) go for 95%…..but if risk is very critical…say hospital industries…. life saving drugs performance….or nuclear industries…..we chose ..or want to take least risk in saying the thing….so we take we cover 99%  p=0.01
    hope this ll give u ..clear pic
     

    0
    #122845

    HF Chris
    Participant

    Try this…pick two different samples (say output from day shift and ouput from second shift). Then graph the lower and upper CL for each sample. Was there really a valid difference? What happens when you adjust the confindence to .95? Look at the spread changes when you change the CI, can you now say there is really a significant difference?
    Chris

    0
    #122846

    Jane Inaura
    Participant

    This is an excellent explanation. My question is  – does the researcher simply choose the P value, or are there statistical studies or data that preceeds identification of which P value to use?  Or, as stated does the researcher simply choose the P value?

    0
    #122847

    Ken Feldman
    Participant

    Let’s not make this more of an issue than it is.  Simply put, the p-value is the calculated probability that the null hypothesis is true.  The concept of p value is linked to your alpha error or Type 1 error.  Your selected alpha is the amount of risk that you are willing to accept with regards to rejecting the null when in fact it is true.  If you think something has happened or that there is a change, you can really cause chaos if in fact nothing has changed and you are reacting to random variation.  If it is costly to think something has happened and it hasn’t, then you don’t want to take a big risk and will probably select an alpha of .01 instead of .05.  If it isn’t a big deal you might select an alpha of .10.  On the other hand, your selected beta error is the amount of risk that you are willing to accept with regards to not rejecting the null when in fact it is false.  If it is costly to miss a change, as in a lab test for cancer, then you want to minimize your chance of missing that change and will select a low beta value.  P value, alpha and beta are all kind of linked together and it is wise to consider them all in developing a testing strategy.  Bottom line is that you don’t select the p value, it is calculated.  You select your alpha and beta risks based on your comfort with being wrong and the cost of being wrong.  Hope that helps.

    0
    #122849

    HF Chris
    Participant

    Darth,
    You point is well made and only one thing that I would like to add. Even if the p value is less than .001, it does not indicate that the results of a study are repeatable or a cure.  It just means that we failed to reject the null hypothesis. Many studies and research have been thrown away because of having a p value less than .05. Finally, we do not select p-value but we do select the CI.
    Chris
     

    0
    #122852

    Ken Feldman
    Participant

    Chris, if you want to get really picky, you don’t pick the CI, you pick the confidence level.  The confidence interval is calculated.  I agree that you accepted/rejected based on some sample.  Does that make it the universal truth, probably not.

    0
    #122858

    HF Chris
    Participant

    Darth,
    Not second guessing you or being picky. The reply was more for the original poster to emphasize what p values are and are not.  I actually like using Signal detection theory calculators to see how many correct hits and misses and incorrect hits and misses you get by being too conservative or too liberal.
    thanks, chris
    Chris

    0
Viewing 9 posts - 1 through 9 (of 9 total)

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