Lack of Randomization

[VRaboy]

I have two data sets of field trials of experimental maize inbreds and hybrids. One trial used a completely randomized design where all the reps for all the treatments were randomized within the nursery. However, for the second field trial the reps were simply planted in numerical order by treatment, ie. no randomization. I want to do statistical analyses to test if there is any significant effect of my experimental treatments. I have used GLM using the mixed model approach and know that that is valid for the first trial where the reps were randomized. But what about the second trial where there was no randomization? 

[PaigeMiller]

One thing you can do is to add a variable into the model indicating the non-random order/sequence and see if that is significant. If it is not significant, then maybe the usual statistical tests are valid (or approximately valid) but I would still be very scrupulous checking residual plots for patterns and clusters and any other odd looking things in the residual plots.

[VRaboy]

Thanks for your prompt reply. That sounds like an excellent approach. I
will work on this in the next couple of weeks (so far I have two
approaches) and after some thought and tests see what looks right.
Victor Raboy