On Tue, 15 Jan 2013 08:22:05 -0800, Michael Sylvester wrote:
In psychological science we require at least a p.05 or better to come to reliable conclusions about the impact of the IV on the DV.
There are too many things wrong with this statement to correct right now but the key point that I think that Prof. Sylvester is trying to make is that one typically sets the Type I error rate alpha= .05, meaning that that an observed inferential statistical test result would occur by chance only 5% of the time if the null hypothesis is true. Onward.
But the flu vaccine only has a p.6 (62%) effectiveness,
Prof. Sylvester has confused vaccine effectiveness with the Type I error rate. Vaccine efficacy and vaccine effectiveness are related concepts but apply to different situations. Quoting from the Wikipedia entry on "Influenza Vaccine": http://en.wikipedia.org/wiki/Influenza_vaccine#Effectiveness_of_vaccine |Effectiveness of vaccine | |A vaccine is assessed by its efficacy, the extent to which it |reduces risk of disease under controlled conditions, and its |effectiveness, the observed reduction in risk after the vaccine |is put into use.[39] One way to view the distinction is that vaccine efficacy is an effect size measure of the vaccine's ability to prevent the development of a case of flu and can be thought of as reflecting the internal validity of the causal relationship between vaccination and illness prevention. Vaccine effectiveness, however, is more like external validity, that is, once one has left the controlled condition of the clinical trial and gone into the community, what percentage of people are NOW prevented from developing the illness. Further quoting from the Wikipedia entry: |In the case of influenza, effectiveness is expected to be lower |than the efficacy because it is measured using the rates of |influenza-like illness, which is not always caused by influenza.[40] |Influenza vaccines generally show high efficacy, as measured |by the antibody production induced in animal models or |vaccinated people,[41] or most rigorously, by immunizing |healthy adult volunteers and then challenging them with virulent |influenza virus.[42] However, studies on the effectiveness of |flu vaccines in the real world are uniquely difficult; vaccines |may be imperfectly matched, virus prevalence varies widely |between years, and influenza is often confused with other |influenza-like illnesses.[43] However, in most years (16 of |the 19 years before 2007), the flu vaccine strains have been |a good match for the circulating strains,[44] and even a |mis-matched vaccine can often provide cross-protection.[45] It should be noted that these concepts come from epidemiology and public health. The CDC provides an online textbook on Epi and an example of how to calculate the efficacy/effectiveness is provided here: http://www.cdc.gov/osels/scientific_edu/ss1978/lesson3/Section6.html Quoting from the CDC website: |Vaccine efficacy/effectiveness (VE) is measured by calculating |the risk of disease among vaccinated and unvaccinated persons |and determining the percentage reduction in risk of disease among |vaccinated persons relative to unvaccinated persons. The CDC reported the vaccine effectiveness in a recent issue of MMWR and the article described how it was calculated can be obtained here: http://www.cdc.gov/mmwr/pdf/wk/mm62e0111.pdf NOTE: the 95% CI is 51%-71%.
so why are we recommending that everyone get a flu shot. With such a low level of significance,could this be the quintessential 'placebo effect' paradigm?
Professor Sylvester has confused apples with mangoes. The Type I error rate has nothing to do with the statistical estimate VE which is essentially a sample statistic estimating a population parameter. As the CDC points out: "the numerator [of VE] (risk among unvaccinated − risk among vaccinated) is sometimes called the risk difference or excess risk." The greater the effect of vaccine in reducing the occurrence of the illness, the greater the difference is. If the vaccine is infective, the difference is small or zero, meaning that the numerator is essentially zero. Dividing zero by any number produces zero which implies that if the vaccine has no effect, VE=0. One could set this as a null hypothesized value, use Type I error rate alpha=.05, and then do a one sample test of the obtained VE against a population VE=0.00.
Any MD on Tips except for Beth's husband?
You want to talk to an epidemiologist and not an MD unless that MD also has a MPH. -Mike Palij New York University [email protected] --- You are currently subscribed to tips as: [email protected]. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=23022 or send a blank email to leave-23022-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
