Page 134 - Breeding and regulatory opportunities, Renaud

Chapter 4

For head weight across all trials in both locations (L, Oregon versus Maine), we

found that at the environmental level Season (four trialing periods) described

the largest portion of the total variance (34%), followed by L with a CV of 18%

).
(Table 4.3Management system (M, Organic versus Conventional) main efect

was small (< 2%), but the three-way interaction (L×S×M) had a CV of about 11%.

Genotype main efect (G) was 13%, and genotype interactions with L and S (14%

and 9%, respectively), were larger sources of variation for head weight than

G×M at 4%. The CV for the four-way interaction G×L×S×M was 11%. This large

interaction was due to trial speciic efects, because attempts to reduce the

complexity of this interaction by ignoring years (so focussing on spring versus

fall) or ignoring season within year (so focussing on year itself) failed (results

not shown).

For days to maturity, the L×S interaction accounted for the largest source of

variation (23%) followed by L and G main efects of 10 and 7%, and the three-

way interaction G×L×S (8%). M main efect and its interaction with G was absent

(0%),andothertwo-andthree-wayinteractionsweresmall.Thelargestsourceof

variation for bead size was G as well as three- and four-way interactions (G×L×S

and G×L×S×M). There was no L main efect for this trait. For the overall quality

rating, sources of variation were distributed among G and L main efects and G,

L and S interactions. For eight of 11 traits analysed the contribution of variation

described by four-way interactions compared to other interaction terms in

the model was relatively large. For this reason, we performed a partitioning

of variance component analysis at each location and season within location

to analyse the contribution of management system to variation at these trial

levels (data not shown).

We performed variance components analyses for the separate locations Maine

and Oregon to more closely examine the partitioning of the variation conditio-

nal on location. At the trial location level (Maine and Oregon), the partitioning

of variance for head weight showed the same pattern as across all trials; trialing

period was important as S contributed to the largest source of variation, followed

by Y×S interactions (data not shown). For other traits such as bead size, again

trialing period efects as S and Y efects were most important. For maturity, the Y

efect in Maine accounted for the largest source of variation, but not in Oregon.

With the higher means for head weight in Oregon (347 g) compared to Maine

116