Figure 2.
Accuracy and precision of estimates of QTL effects across environments obtained by the general and approximated single-QTL models applied to single-QTL data. The graph represents estimated average trend and a sampled 95% vicinity of the *p*-dimensional point {*a*_{i}/σ_{i}, i = 1,… *p*} from 200 Monte-Carlo runs for situation *S*_{2} (described in Table 1). Open circles and light region are for the general model MG, black circles and shaded region are for the approximated model MA. The regions were obtained as follows. Let *a*_{i} and σ_{i} be the QTL effect and the residual standard variation at the *i*th environment (*i*…., *p*), and *a*_{ij} and σ_{ij} be the corresponding estimates at *j*th run (*j* = 1,…, 200). As a measure of discrepancy of the estimated parameters from the expected, across environments, we used the index
The sampled 95% volume is defined as a *p*-dimensional sphere, *S*_{0.95}, of minimal radius with center at point {*a*_{i}/σ_{i}, i = 1,…, *p*}, which includes 95% of estimated results *r*_{j} = (*a*_{ij}/σ_{ij}, i = 1,…, *p*) from the simulations. The sphere provides a 95% region of estimated curves *a*_{j}(μ(*E*)) in the plane of the estimated effects {*a*_{ij},μ(*E*_{i})}:
According to our calculations, the radius of *S*_{0.95} is smaller for MA than MG (0.672 *vs.* 0.805).