[R] Combining means

[Holger Danner]

Hi volunteer statisticians,
I want to combine the stdev of 6 populations of means for two different 
treatments and there seem to be many methods for doing so.
It is QPCR data consisting of 3 technical replicates building one mean 
and 6 biological samples in two treatment groups. Now as an overview I 
just assumed each value to be a replicate it self within each of the two 
treatments and calculated an overall mean with overall stdev. But this 
seems to be a bit of a mistake. Is it?
First - I think I need a more robust mean. So I applied the geometric 
mean. Next I combined the deviations by calculating with the formula for 
error propagation (wiki german "Fehlerfortpflanzung"). Isn't this a too 
complicated way?

So:

3 technical replicates build one mean with stdev
6 of these build the population of one treatment
2 treatments....

Any suggestions which method I should use and how I perform the 
calculations?

Thanks for your help
Best Holger