home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
gimli.bioss.sari.ac.uk
/
gimli.bioss.sari.ac.uk.zip
/
gimli.bioss.sari.ac.uk
/
pub
/
courses
/
ExptDesignAnalysis
/
sessnew4a
< prev
Wrap
Internet Message Format
|
1995-02-16
|
4KB
From kar@uk.ac.sari.rri Wed Apr 3 09:43:44 1991
Received: from rri.sari.ac.uk by sass.sari.ac.uk; Wed, 3 Apr 91 09:43:43 BST
From: Karen Robertson <kar@uk.ac.sari.rri>
Date: Wed, 3 Apr 91 09:40:29 BST
Message-Id: <5684.9104030840@rri.sari.ac.uk>
Via: rri7.yprri; Wed, 3 Apr 91 09:40:29 BST
To: graham@uk.ac.sari.sass
Subject: sessnew4a
Status: RO
"
PRACTICAL SESSION 4
In this session nine treatments are applied to four blocks of
nine units, each of which is divided into three sub-blocks (or
main-plots) of size three. The treatments have a factorial
structure A*B each factor having three levels. In Example ED4.1
the confounding is restricted to the interaction between A and B.
In Example ED4.2 it is distributed across all treatment effects. In
Example ED4.3 a lattice design for nine treatments based on Example
ED4.2 is analysed.
**** Design ED4.1 Confounded 3 x 3 factorial (Plan A4.1) ****
In this design half the information on the interaction between factors
A, B is confounded with the variation between sub-blocks.
Define the treatment factors and read the basic data.
"
unit[nval=36] unitno
factor[levels=9] T
factor[levels=3] A,B
open 'sess4.dat1'; channel=2; filetype=input
read [print=*; channel=2] unitno, T,A,B,TE,PE
"
Generate the blocking factors
"
factor [levels=4; values=9(1...4)] block
factor [levels=3; values=3(1...3)4] subblock
factor [levels=3; values=(1...3)12] plot
"
!!! The next stage is non-standard - randomise the allocation and create
the variate to be analysed.
"
random [block/subblock/plot; seed=12345] T,A,B,TE
calculate [print=summ] V1 = TE + PE
"
End of non-standard stage !!!
"
"
Print data
"
print unitno,block,subblock,plot,T,A,B,V1; field=4(9),3(5),8; decim=7(0),3
"
Analyse the data
"
blocks block/subblock/plot
treatment A*B
anova V1
close 2
"
**** Design ED4.2 Balanced 3 x 3 factorial (Plan A4.2) ****
In this design both main effects and interactions are confounded with
blocks.
"
open 'sess4.dat2'; channel=2; filetype=input
read [print=*; channel=2] unitno,T,A,B,TE,PE
"
Use same block factors as for Design ED4.1
!!! The next stage is non-standard - randomise the allocation and create
the variate to be analysed.
"
random [block/subblock/plot; seed=12345] T,A,B,TE
calculate [print=summ] V2 = TE + PE
"
End of non-standard stage !!!
Print data
"
print unitno,block,subblock,plot,T,A,B,V2; field=4(7),3(5),8; decim=7(0),3
"
Analyse the data
"
blocks block/subblock/plot
treatment A*B
anova V2
"
**** Design ED4.3 Balanced Factorial Design (Plan A4.2) ****
This design is identical to the above except that the nine treatments
have not got factorial structure. A pseudo=factorial structure based
on A, B is imposed. The analysis is the same but the last two lines
become (i.e. you need now enter only) :
"
treatment T // (A*B)
anova V2
"
Exercises Session 4
Exercise 4.1 Data file 'sess4.dat3' contains data for a randomised
block design of 4 blocks of 9 treatments. Read
the data through
open 'sess4.dat3'; channel=3; filetype=input
read [print=*; channel=3] unitno, T, A,B,V
use the following instruction to create a two-way table
of means.
tabulate [print=m; class=block,T; margin=yes] V
Confirm that square-root (V) appears to have more
stable variance and analyse.
Exercise 4.2 Often when the analysis of Exercise 4.1 is appropriate
an alternative analysis is given by a 'generalised
linear model'. Replace the analysis of variance
directives with
model [distrib=poisson; dispersion=*] V
terms block + T
fit block + T
predict [print=d,p,se] T
"
stop