Agriculture Reference
In-Depth Information
Reject
Select
Reject
Select
Select
Select
Reject
Reject
Independent Culling
Independent Culling
Reject
Select
Select
Reject
Index
Index
Figure 7.4
Association between independent culling and
index selection when there is high correlation between two
selectable traits.
Figure 7.5
Association between independent culling and
index selection when there is low correlation between two
selectable traits.
indices. In simple terms, however, variate or character
values in a selection index can be weighted either by:
although all called analysis, are in fact statistical
transformations which produce various equations of
multi-variate data, usually with minimum correlation
between traits or maximum discrimination between
genotypes. The problem with statistical weights is
again that they will be different from data set to data
set. In some cases the weights do indeed show some
biological meaning but in other cases there appears
to be no coherent association.
•
Economics
where the potential economic impact
of each trait is estimated and the datum recorded
of each variate expression is weighted by that value.
For example, the average price paid per unit weight
can usually be predicted from past seasons and an
increase in productivity could be related in money
terms by an appropriate weight. Similarly if a par-
ticular insecticide costs a unit more per acre than if
biological resistance is incorporated then that resis-
tance will accrue monetary value. The problem with
economic weights is that they change from year to
year. If there is over-production of a product in any
year then there is a tendency for the unit weight price
to drop etc.
Selection indices can be extremely useful in plant
breeding and their true value is perhaps yet to be real-
ized. If index selection is carried out in a meaningful
manner, then index selection should be more effective
than independent culling.
Errors in selection
•
Statistical features
where the weight values are
derived according to some statistical procedure. The
most commonly used routines have involved
multi-
variate transformations
such as principal component
analysis, canonical analysis or discriminant func-
tion analysis. Each of these statistical techniques,
Each time selection is applied there is a chance that an
error will occur. Errors in selection happen because the
true genotype value is masked by environmental effects
or because of administrative or clerical error.
