No announcement yet.

Re: Stats Power. Report Confidence Limits - p values

This topic is closed.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Re: Stats Power. Report Confidence Limits - p values

    Statisticians correct if I'm wrong but in this example you cannot try to
    establish a significant difference with an 80% of confidence. In a student
    test you compare average values of two groups. Depending on the differences
    between this values, their standard deviation and the sample size you can
    establish a statistically significant difference (usually people set alpha
    to 5% meaning that there is only 5% probability that if a difference is
    found is just due to chance alone).

    In this example you have to establish first the groups you want to establish
    a significant difference: My guess TMJ through plastic splint and non
    treatment (since this 80% is solely based on success using this method in
    relation to failure, and not related to other methods)

    Even if TMJ through plastic splint was only 1% successful for a large enough
    sample size it would be statistically significant to non treatment 0%
    (success) with basically a 100% confidence (this assuming that the variation
    of these percentages is 0 or close to 0, if not then to get closer to 100%
    confidence you'll need a larger sample size)

    Same thing if you are comparing two methods with say 60% and 80% success
    rate and having a large enough sample size (that accurately describe this
    percentages) you can also establish a statistically significant difference.
    Again if the variation of these percentages is small and the sample size
    large enough you can achieve a percentage of confidence close to 100%.


    ----- Original Message -----
    From: "Bruce Etnyre, Ph.D., P.T."
    Sent: Wednesday, January 26, 2005 11:34 AM
    Subject: Re: [BIOMCH-L] Stats Power. Report Confidence Limits - p values

    > Would anyone like to comment on clinical significance?
    > For example, my dentist can completely cure temporomandibular joint
    > dysfunction (TMJ) in 80% of patients by applying a plastic splint to
    > the teeth. Scientifically, this is not statistically significant. It
    > does not even approach statistical significance. So statistically we
    > would accept the null hypothesis that there is no difference in the
    > treatment of TMJ by using this device or not because the level of
    > confidence is only 80%.
    > Statistically, my dentist is a failure, but clinically, a world-beater!
    > BE
    > At 9:58 AM -0500 1/26/05, Dr. Chris Kirtley wrote:
    > >This brings up an interesting issue. I spoke to a statistician last year
    > >told me that the only reason Fischer came up with these tables, and
    indeed the
    > >whole concept of significance testing, was because there were no
    > >available at that time. He considered T-tests, ANOVAs etc. to be stopgap
    > >methods that could to be used to get an approximate estimate until better
    > >computing power came along (as he expected it would).
    > >
    > >If Fischer were to be alive today, he would likely be appalled that we
    > >still using his extremely simplified methods. Apparently any decent real
    > >statistician worth his salt these days performs a simulation in order to
    > >compute the likelihood of error. I was never able to find out how
    > >this is done,
    > >but perhaps someone else on the list can enlighten us? It really is
    > >time all of
    > >us in biomechanics moved into the modern age!
    > >
    > >Bryan Kirking wrote:
    > >
    > >> To comment and question some of Dr. Allison's insight:
    > >>
    > >> >>My understanding of the arbitrary "line in the sand" of 0.05 was
    > >> >>originally due to the choice of the original tables (pre computer)
    > >
    > >--
    > >Dr. Chris Kirtley MD PhD
    > >Associate Professor
    > >Dept. of Biomedical Engineering
    > >Catholic University of America
    > >Washington DC 20064
    > >Alternative email:
    > >
    > >-----------------------------------------------------------------
    > >To unsubscribe send SIGNOFF BIOMCH-L to
    > >For information and archives:
    > >-----------------------------------------------------------------
    > --
    > Bruce Etnyre, Ph.D., P.T.
    > Kinesiology Department
    > Professor and Chair
    > Rice University
    > 6100 Main MS 545
    > Houston, Texas 77005
    > USA
    > Phone: (713)348-5936 or 8816
    > FAX713)348-8808
    > -----------------------------------------------------------------
    > To unsubscribe send SIGNOFF BIOMCH-L to
    > For information and archives:
    > -----------------------------------------------------------------

    To unsubscribe send SIGNOFF BIOMCH-L to
    For information and archives: