home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!uvaarpa!darwin.sura.net!wupost!bu.edu!jade.tufts.edu!news.tufts.edu!sage.hnrc.tufts.edu!jerry
- From: jerry@nutmeg.hnrc.tufts.edu (Jerry Dallal)
- Newsgroups: bit.listserv.stat-l
- Subject: Re: Meta Analysis ???
- Message-ID: <1992Aug29.184250.330@nutmeg.hnrc.tufts.edu>
- Date: 29 Aug 92 23:42:49 GMT
- References: <STAT-L%92082805302967@VM1.MCGILL.CA>
- Organization: USDA HNRC at Tufts University
- Lines: 23
-
- In article <STAT-L%92082805302967@VM1.MCGILL.CA>, LOFJUST@VM.UNI-C.DK (Just Jensen) writes:
- > A colleauge has asked how to combine information from different experiments
- > all related to various treatments with hormones. These are all very expensive
- > experiments and therefore tends to be small. What is the best way to
- > combine evidence?. I have suggested ML/Weithted Least square.
- > In the back of my head i have the word META-Analysis.
- > I would very much appreciate references and comments on this/these? subjects
- >
-
- I think people on the net would like to help everyone as much as possible,
- but pointing you to a set of references regarding meta-analysis would be doing
- you more of a disservice than a service. If you've got sparse, expensive data
- I suggest you hire a statistician having experience with such data.
- If the data don't pass the inter-ocular traumatic test (i.e., the results don't
- hit you between the eyes) there's a good chance you'll get into big trouble if
- you try to solve your problems with some technique with which you have no
- experience.
-
- Certainly, if the data cost thousands to collect, they're worth a few hundred
- to get properly assessed.
-
- (This is *not* a flame. It is offered as a helpful suggestion and my
- professional assessment of your predicament.)
-