2^4 On-farm trial - only 4 plots per farm -> what design possibilities?


I have to design an on-farm trial with possibly about 600 farmers involved (setting: Madagascar, so rather illiterate participants).

There are 4 factors, each at two levels (possibly one at 3 levels but that might get too complicated). Each farmer should have a maximum of four plots, and each farmer should have a control plot, no replicates within farmers, so farmers are blocks. Farmers also have a choice which technique (factor) they are interested in, so they shouldn't be too much "forced" to try weird treatment combinations that they are not interested in. We expect some degree of interaction among factors.
I've been looking around, it get's pretty statistical - fractional factorial design, confounding, motherXbaby design (which seems more used and appropriate in onefactorial trials?), possibly stepwise design which I found interesting but not much literature other then Mutsaers (1990 and 1997), or main effects trial combined with (fractional) factorial (Gomez - Statistical Procedures for Agricultural Research).

My stats background is not impressive enough that i can follow throughmost of these methods descriptions in the literature, just to find out whether that might work for me or not.

So, generally, can somebody give me at least some directions which way I could go and which way not, or somebody has some experience?
(I can give some more details if necessary)

Susi :cool:
Have you finished this design? If so, it would be interesting to know what you have chosen?

Or otherwise, do you still want to discuss the design issues?