Hi Brian et al,
I am not a biostatistician but have spoken to several about Bayesian
analysis. My sense is that the information provided about some of the
priors is fairly limited ie. "loosely informative inverse chi-square prior
distributions for all the variances".
I can't recall who mentioned the PLAC I study but the same comment applies
to the REGRESS study. One has to wonder about the point estimate for that
study, 0.67, which is further from the null (1.0) than the pooled
estimate, 0.70 (figure 3).
Thus bayesian hierarchical analysis is not necessarily incorrect but
something unorthodox is happening in this analysis. I would think that
peer review would catch this, or perhaps it speaks to a lack of biostats
perspective there...Dean
Dean Giustini, UBC biomedical branch librarian
Diamond Health Care Centre and Vancouver Hospital
Vancouver BC, Canada V5Z 1M9
blog: http://weblogs.elearning.ubc.ca/googlescholar
wiki: http://hlwiki.slais.ubc.ca
On Wed, 2 Jul 2008, Brian Alper MD wrote:
> To recap some of this message thread:
>
> The article citation is Afilalo J, Duque G, Steele R, Jukema JW, de
> Craen AJ, Eisenberg MJ. Statins for secondary prevention in elderly
> patients: a hierarchical bayesian meta- analysis. J Am Coll Cardiol.
> 2008 Jan 1;51(1):37-45. PMID 18174034
>
> A good rule of thumb is to evaluate how the data was selected and
> collected for critical appraisal before concern regarding the
> statistics. That was done in this case (and that is a daily process in
> the DynaMed workflow). This particular systematic review met the
> traditional quality criteria for these concepts.
>
> So we are faced with a "face validity" concern when trying to interpret
> the statistics for this particular meta-analysis. There are numerous
> individual data comparisons that look like they should not be
> statistically significant at all, yet are all showing similar results
> for "posterior median relative risk (95% credible interval)" in the
> Figure 2: Bayesian forest plot for all-cause mortality which is the
> primary figure representing a key finding in this meta-analysis.
>
> One of the authors of this paper is a statistician, but is that
> sufficient to accept this statistical data? I'd be curious if some of
> you could look at the actual paper and comment on the statistics in this
> case.
>
> Thanks.
>
> Brian
> ---------------------------
> Brian S. Alper, MD, MSPH
> Editor-in-Chief, DynaMed (www.DynamicMedical.com)
> Medical Director, EBSCO Publishing
> 10 Estes St.
> Ipswich, MA 01938
> office (978) 356-6500 ext 2749
> cell (978) 804-8719
> fax (978) 356-6565
> home (978) 356-3266
> "It only takes a pebble to start an avalanche."
>
> This e-mail and any attached files transmitted are confidential and
> solely for the use of the intended recipient. It may contain
> information which is covered by professional or other privilege. If you
> are neither the intended recipient of this e-mail nor the person
> responsible for delivering it to the intended recipient, be advised that
> you have received this e-mail in error and that any use of it is
> strictly prohibited. Please notify the sender immediately be reply
> e-mail and then delete from your system. EBSCO Industries, Inc., its
> subsidiaries and divisions, accept no responsibility for any loss or
> damage suffered by any person arising from the use of this e-mail.
>
|