> Similarly, it seems strange to me to manage crops
> conventionally in fields that have a long history
> of sustainable management.
Agreed. The interaction of variety with management factors might have to be
looked at across locations making formal statistical tests difficult. But
fuzzy information is usually better than none, and in the worst-case the
results (OP vs Hybrid) would simply be applicable separately for each
> If you set up randomized and replicated plots in a
> sustainably managed field where you've just killed
> a nice cover crop without herbicides, are you going
> to go back in and use herbicides and sidedress N on
> the conventional plots? Doesn't make sense.
This is not an uncommon kind of problem in multi-location trials. There are
at least four ways to handle it.
1. Treat it as a split-plot, but have the "main-plots" (management system)
at different sites, grouping the sites (before execution) into pairs on the
basis of accessory factors such as region or geomorphology. This is
analogous to pairing of individuals in clinical trials.
2. Shotgun the trial (without management factor included) at many
locations, and figure out why locations behaved differently after execution
of the experiment (dangerous but sometimes necessary). If the locations
don't interact with the treatment (behave differently), then the problem
disappears. For example, I wouldn't be surprised if an adapted hybrid
uniformly outperformed the OP at all locations, at least if the stand were
28,000/A or so.
3. Develop large, long-term management plots within location and use these
as main-plots in the study. Expensive, but we might be able to plug into
some existing sites where this has already been done.
4. Identify certain dominant factors associated with management (such as N
rate) and apply these as treatments in the experiment, regardless of
location, as stand-ins for management system. This is reductionist, but if
done carefully, can yield useful information. One problem is that in
complex systems, applicable factors are numerous, and these experiments can
grow very, very large to accomodate conceivable interactions.
Anyhow, designing experiments like this is a can of worms, but kind of fun
if you have the right perspective, like finding the right lure to catch a
To Unsubscribe: Email email@example.com with "unsubscribe sanet-mg".
To Subscribe to Digest: Email firstname.lastname@example.org with the command