From: Eva Hogervorst (eva.hogervorst@linacre.oxford.ac.uk)
Date: Thu 30 May 2002 - 01:44:20 EST
Message-id: <E17D5ca-0000xg-00@wing2.herald.ox.ac.uk>
Dear Sir/Madam
I have a question about the simulated normal population.
I am writing a review for Cochrane which involves meta-analyses.
For these I need the mean difference and SD between the pre and posttreatment
performance of drugs vs placebo. However, often these data are not given
and it is usually not possible to get the real data from the authors.
Calculating the mean difference = post-pretreatment performance
Howver the SD of that mean difference is calculated as
the square root of the sum of the variance of pre and posttreatment performance
As I have the actual data of two studies, this allowed me to make
comparisons. The estimated SD (using the formula) often overestimates
the real SD sometimes by more than a factor 3. This varied whether the
SD was small (overestimation less) or larger (overestimation went up).
So I thought that I could use the simulated normal population
in R and put in the mean and SD and N to then calculate the mean
difference and SD.
I used only 3 data points but the factor by which the simulated SD
overestimated the actual SD seemed constant at 2.2 which at least is
an improvement over the first formula.
This is a serious problem as it increases type II errors. Do
you have any idea how I can solve it ?
Many regards
Dr. E. Hogervorst
University of Oxford, OPTIMA
Radcliffe Infirmary Trust,
Woodstock Rd, Oxford OX26HE
tel 00 44 (0) 1865 228512
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
This archive was generated by hypermail 2.1.3 : Wed 16 Oct 2002 - 11:57:18 EST