Agriculture Reference
In-Depth Information
Table 6.5
Average plant height of each of the 45 F
1
's and the 10 parents in a half diallel with selfs.
Global
Global
328
Helios
Helios
341
352
Jaguar
Jaguar
336
310
263
Starr
Starr
329
271
293
287
93.C.3
93.C.3
269
308
271
312
292
Westar
Westar
256
350
279
299
324
273
DNK.89.213
DNK.89.213
284
313
290
266
259
270
293
Cyclone
Cyclone
321
263
280
241
285
243
273
201
Hero
Hero
246
261
315
261
241
256
250
244
231
Reston
Reston
306
327
295
287
284
296
275
265
248
277
GCA
+
18.4
+
26.4
+
9.4
+
0.4
+
0.4
+
0.4
−
6.6
−
22.6
−
28.4
+
2.4
In open-pollinated species, where GCA is the only
parameter of interest, then it has been suggested that
other designs such as topcross or polycross would yield
equally reliable results with less effort and that these
alternative methods provide the opportunity to test
many more parental lines. Similarly it has been argued
that in many instances North Carolina I designs (where
a set of
p
parents to be tested are each inter-crossed with
a set number of other parents and where each parent
under test is not necessarily crossed to the same tester)
or North Carolina II designs (where a set of
p
parents
are crossed to a common set of
n
different parents and
where each parent under test is crossed to the same set of
non-test parental (or tester) lines) would offer a better
alternative to diallel designs and Griffing's Analysis.
Many studies have shown that the GCA values of
parents from diallel analyses are similar to actual pheno-
typic performance of the parents. It has, therefore, been
argued that it is not necessary to progeny test potential
parents in a plant breeding programme but simply to
'
cross the best with the best
'. Many practical plant breed-
ers often add to this statement, however, '
cross the best
with the best, and hope for the best
', but perhaps that is
what we would be doing anyhow.
rapeseed cultivars while the others are canola (edible)
types. Crossing resulted in
n
[
−
]
/
=
45 differ-
ent F
1
families. Over the following winter each of the
45 F
1
families were grown in a two replicate randomized
complete block design which also included the 10 par-
ent selfs making a design with 55 entries
n
1
2
(
[
+
]
/
)
n
n
1
2
and two replicates (i.e. 110 plots).
Throughout the growth of this experiment a number
of different traits were recorded on each of the 110 plots.
To avoid excessive repetition we will only consider one
of these characters, plant height at end of flowering.
The average plant height of each of the 45 F
1
s and
the 10 parents are shown in Table 6.5. The data used
were the sum of two plant heights (cms) as two readings
were made on each of the replicate plots.
From the data the total variance (sum of squares) is
partitioned into differences between the two replicate
blocks (Reps), general combining ability, specific com-
bining ability and an error term (based on interactions
between replicates and other factors). Sum of squares
(SS) and mean squares (MS) obtained are shown in
Table 6.6.
The basic assumption of this experiment was that the
ten parental lines were
chosen
as representative of the
wide range of
B. napus
cultivar types that were avail-
able. We are therefore analyzing a
fixed effect model
and
all the mean squares in the analysis are tested for signif-
icance (using the 'F' test) against the error mean square
(i.e. 1545).
From the analysis the overall replicate block effect
(i.e. difference between replicate one and replicate two)
was not significant. An F-value is obtained for specific
Example of Griffing analysis of half diallel
Let us consider now an example of a half diallel. A half
diallel crossing design between ten homozygous lines
of spring canola (
Brassica napus
) was carried out in the
spring of 1992. The parental lines were: Global, Helios,
Jaguar, Starr, 93.C.3.1, Westar, DNK.89.213, Cyclone,
Hero and Reston. Hero and Reston are both industrial
