Before using the
REML method to estimate genetic variance components we shall rerun
the analysis used in Case Study 3
to compare weaning weights of lambs. First we must disregard the
lambs for which the response variable weaning weight was not
recorded. This can be achieved by using the GenStat Spread →
Restrict/Filter By Value... command and excluding missing values (*) for weaning weight
This time we shall alter the way that
breed genotypes are defined in the least squares analysis. Instead
of referring to the breeds by their genotype, D X D, D X R, R X D
and R X R we shall consider separate effects for ram breed, ewe
breed and their interaction, and re-parameterise the model
accordingly. We can run this model both by least squares analysis
of variance and by REML.
Let us first consider the least squares
approach. Using Stats
Regression Analysis
Generalized Linear
Models and completing the dialog box as shown and clicking the Options... button and then ticking 'Accumulated', we obtain the analysis of variance indicating that the breed of ram x breed of ewe is insignificant (variance ratio = 0.15).
|
|
***** Regression Analysis *****
Response variate: WEANWT
Fitted terms: Constant + YEAR + SEX + AGEWEAN + DL + DQ + RAM_BRD +
EWE_BRD + RAM_BRD.EWE_BRD
**Accumulated analysis of variance**
Change |
d.f.
|
s.s.
|
m.s.
|
v.r.
|
+
YEAR |
5
|
1208.149
|
241.630
|
48.92
|
+
SEX |
1
|
55.983
|
55.983
|
11.34
|
+
AGEWEAN |
1
|
344.206
|
344.206
|
69.69
|
+
DL |
1
|
151.513
|
151.513
|
30.68
|
+
DQ |
1
|
275.795
|
275.795
|
55.84
|
+
RAM_BRD |
1
|
44.881
|
44.881
|
9.09
|
+
EWE_BRD |
1
|
30.223
|
30.223
|
6.12
|
+RAM_BRD. EWE_BRD |
1
|
0.754
|
0.754
|
0.15
|
Residual |
687
|
3392.947
|
4.939
|
|
Total |
699 |
5504.450
|
7.875
|
|
|
|