Beruflich Dokumente
Kultur Dokumente
Total
9
6
66
8
6
49
9
7
63
5
4
45
31
23
223
Fuentes var.
g.l.
S.C.
C.M.
Carnes
31.9
10.6
5.53 (3.27)
Bloques
58.5
6.5
3.39 (9.27)
Error
27
51.4
1.9 (s2)
Total
39
141.8
DISEO ANIDADO/JERRQUICO
En cada unidad experimental se hacen varias
observaciones. Estas observaciones no se
consideran repeticiones sino SUBMUESTRAS
(no son observaciones independientes).
Ejemplo: Degustacin de carne. Idem a diseo
completamente al azar (40 jueces), pero cada
juez valora tres veces cada carne:
DISEO ANIDADO/JERRQUICO
Carne 1
Carne 2
Carne 3
Carne 4
Total
(8,9,10)
(7,8,9)
(8,9,10)
(4,5,6)
93
27
24
27
15
(5,6,7)
(3,4,5)
18
Total
198
69
12
147
189
135
669
Fuentes
g.l.
SC
CM
Carnes
95.7
31.9
3.48 (3.36)
36
329.7
9.1
9.16(36.80)
80
80.0
1.0
1982
1352 669 2
SC carnes :
...
95.7
30
30
120
27 2
12 2 1982
1352
329.7
SC error exp. :
...
...
3 30
30
3
27 2
SC error mues. : (8 9 10
) ... 80
3
2
2 NIVELES DE TRATAMIENTOS
PADRE (k)
1
2
MADRE (r)
HIJOS (n)
4,2
6,4
10
8,6
14
3,9
12
12,10
22
9,11
20
X ...
X ...
30
84
54
2 NIVELES DE TRATAMIENTOS
Fuente
Padres (k)
Madres (r)
Progenie (n)
Total
g.l.
1
(k-1)
4
k(r-1)
6
kr(n-1)
11
krn-1
SC
CM
48
48
4.36 (1.4)
44
11
2.36 (4.6)
28
4.7
120
30 2 54 2 84 2
SC padres :
48
6
6
12
6 2 10 2 14 2 30 2
SC madres :
... 44
2
2 6
2
62
2
2
SC progenie (4 2 ) ... 28
2
2
84
SC Total (4 2 ... 112 )
120
12
3 NIVELES DE TRATAMIENTOS
In the following study glycogen content was evaluated in rats after fed three
different diets (treatments). Duplicate reading were made on each of three
preparations of rat livers from each of two rats for three different
treatments. (Note: this is a 3-level nested ANOVA.)
Treatments (a=3)
Rats (b=2)
Preparations (c=3)
Readings (n=2)
2
3
1
3
2
3
1
3
2
3
131 131 136 150 140 160 157 154 147 151 147 162 134 138 135 138 139 134
130 125 142 148 143 150 145 142 153 155 147 152 125 138 136 140 138 127
proc glm;
class treat rat prep;
model gli=treat rat (treat) prep (rat treat);
Test h=treat e=rat (treat);
Test h=rat(treat) e=prep(treat rat);
Means treat rat(treat) prep (treat rat); Run;
proc mixed;
class treat rat prep;
model gli=treat;
random rat(treat) prep(treat rat);
lsmeans treat; run;
NESTED
PROC
MIXED
PROC GLM
proc sort;
by treat rat;
proc nested;
class treat rat prep;
var gli;
run;
bST treatments1
DMI,kg/d 3.5%FCM,
kg/d
Milk,
kg/d
Period
1CO
1
2
3
4
1
2
3
4
1
2
3
4
CO
Med
31.64
28.08
22.77
17.52
32.18
28.78
23.66
19.47
21.15
21.22
20.06
18.26
DI
SE
0.80
1.05
1.43
1.46
0.96
1.04
1.34
1.51
0.63
0.60
0.62
0.69
Med
33.11
29.53
25.15
19.77
33.33
30.23
26.87
22.24
20.10
20.59
21.18
19.69
SR
SE
0.75
0.98
1.34
1.37
0.90
0.97
1.26
1.43
0.59
0.56
0.58
0.64
Med
32.22
29.51
25.93
22.07
32.43
29.45
26.76
23.48
19.98
21.24
21.06
20.44
SE
0.81
1.05
1.43
1.48
0.96
1.04
1.35
1.53
0.64
0.60
0.63
0.70
CO
vs.
DI
NS
NS
NS
NS
NS
*
NS
Contrast
CO
DI
vs.
vs.
SR
SR
NS3
NS
NS
NS
*
NS
**
NS
NS
NS
NS
NS
*
NS
**
NS
NS
NS
NS
NS
NS
**
NS
Nonpregnant
Rat#
1
2
3
4
5
6
Avg
7
8
9
10
11
12
Avg
8-12
7.5
10.6
12.4
11.5
8.3
9.2
9.92
13.3
10.7
12.5
8.4
9.4
11.3
10.93
20-24
0.8
1.6
5.6
7.5
0.5
3.8
3.30
11.1
9.3
10.1
5.7
3.8
8.5
8.08
ORDENES SAS-MIXED/RANDOM
(CS: las correlaciones no dependen del tiempo)
proc mixed;
class trat udexp t;
model y=trat t trat*t;
Random udexp;
lsmeans trat/pdiff;
Run;
(Variables dependientes 1 x 1)
582
277
277
277 582
277 277
1
0.48
0.48
1
0.48 0.48
277
582
0.48
0.48
1
3
1
2
3
VAR T1 T2 T3
COV 1-2, 1-3
MATRIZ
CORRELACIONES
0.48 = 277/582
2 (ti)
CS
Simtrica compuesta
cte
cte
CSH
Simtrica compuesta
heterognea
cte
AR(1)
Autoregresiva
orden 1
cte
ARH (1)
Autoregresiva
orden 1
heterognea
UN
No estructurada
566
260
120
260 566
120 260
1
0.46
0.46
1
0.21 0.46
260
566
0.21
0.46
1
3
1
2
3
VAR T1 T2 T3
COV 1-2, 1-3
MATRIZ
CORRELACIONES
0.46 = 260/566 =
120/260
0.21 = 120/566
COL 1
COL 2
COL 3
537
317
225
317
786
272
225
272
396
COMPARACIN DE MODELOS
Se dan rdenes para distintos modelos con los
mismos datos
Salida (por defecto)
AIC Criterio de informacin de Akaike
BIC Criterio de informacin de Schwarz (+severo)
(elegir modelo que de ms bajo en valor absoluto)
El mejor modelo puede variar s/var dependiente
No siempre se cumple el criterio de convergencia
RDENES SAS-MIXED/REPEATED
proc mixed;
class trat udexp t;
model y=trat t metodo*t;
repeated/type=cs sub=udexp(trat);
lsmeans trat t trat*t ;
Run;
(Variables dependientes y estructuras 1 x 1)
PROC MIXED-OPTIONS
Constrastes polinomiales:
class trat;
model y=trat | t | t | t/htype=1;
Results, with the exception of data pertaining to SeMet and SeCys, were analyzed
using the MIXED procedure (SAS Inst. Inc., Cary, NC) according to Littell et al.
(1998). Sources of variation included treatment, time, and the treatment time
interaction. The random variable was horse within treatment. Preexperimental data
were used for covariate adjustment where appropriate. The PCV values were used
as a covariate in the analysis of total Se in whole blood. Each variable analyzed was
subjected to 3 covariance structures: autoregressive order, compound symmetry,
and spatial power1(Littell et al., 1998). Using the largest Akaike information
criterion and Schwarz Bayesian criterion, the spatial power was the covariance
structure that fitted the model best. Results are presented in tables as least squares
means and SED. A comparison was made between treatments SY03 (Se yeast) and
SS03 (selenite) using the PDIFF option of SAS, because they had similar
concentrations but different sources of Se. The dose response to concentration of
Se yeast in the diet was considered for treatments SY02, SY03, and SY04. If a
significant interaction was detected (P < 0.05), pairwise comparisons of treatment
means at each sampling date and between sampling date within treatment were
made using the PDIFF option of SAS.
1type
= sp(pow)
DISEO SPLIT-PLOT
Es un diseo muy utilizado en agricultura. A veces en
experimentos factoriales, un factor requiere un tamao
de unidades experimentales distinto del otro (s).
Ej. Factor 1= abonado de praderas requiere parcelas
grandes (por ej. 1 ha=main plots), pero Factor 2 (por ej.
Pienso) se aplica a cada vaca (por ej. 3 vacas/ha = splitplot).
MODELO
Las parcelas de abonado se asignan al azar dentro de los main plots
(12 parcelas de 1 ha, 4 dosis de abonado, 3 repeticiones)
A2
A1
A1
A3
A4
A3
A4
A1
A4
A23
A2
A2
SPLIT-PLOT
Yijk i j k ij ( ) jk ijk
SE DE LAS DIFERENCIAS
Al variar el tamao de las unidades experimentales, es necesario calcular
varios SE para las diferentes comparaciones entre medias del modelo.
S2a = cuadrado medio del error (main plots)
S2b = cuadrado medio del error (sub plots)
r = nmero de bloques
A = nmero de niveles factor A (main plots)
B= nmero de niveles factor b (sub plots)
2Sa2
2S2b
4Sa2
(rb)
(ra)
SPLIT-PLOT
Parcela
Tipo de pasto
Suplementacin mineral
Produccin de leche, kg
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
4
4
1
1
2
2
3
3
2
2
1
1
4
4
3
3
1
1
2
2
4
4
3
3
2
1
2
1
1
2
2
1
1
2
2
1
1
2
1
2
2
1
1
2
2
1
1
2
30
29
27
25
26
28
26
24
32
37
30
31
34
37
33
32
34
31
30
31
36
38
33
32
VS
MIXED
proc mixed;
class parcela pasto suplem;
model leche=pasto suplem pasto*suplem;
random parcela parcela*pasto;
lsmeans pasto suplem pasto*suplem; run;
DOUBLE BLOCKING
(T.R. MORRIS)
DOUBLE BLOCKING-ANOVA
ANOVA B
ANOVA A
Source
d.f.
Source
d.f.
Breeds
Blocks within breeds
Diets
Breed x diet
Error
Total
1
6
2
2
12
23
Blocks
Breeds
Diets
Breed x diet
Error
Total
3
1
2
2
15
23
ANOVA C
Source
d.f.
Blocks
Breeds
Error (a)
Main plots
Diets
Breed x diet
Error (b)
Sub-plots
3
1
3
7
2
2
12
23