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
/
sessnew2a
< prev
next >
Wrap
Internet Message Format
|
1995-02-16
|
5KB
From kar@uk.ac.sari.rri Wed Apr 3 09:43:23 1991
Received: from rri.sari.ac.uk by sass.sari.ac.uk; Wed, 3 Apr 91 09:43:23 BST
From: Karen Robertson <kar@uk.ac.sari.rri>
Date: Wed, 3 Apr 91 09:40:07 BST
Message-Id: <5655.9104030840@rri.sari.ac.uk>
Via: rri7.yprri; Wed, 3 Apr 91 09:40:07 BST
To: graham@uk.ac.sari.sass
Subject: sessnew2a
Status: RO
"
PRACTICAL SESSION 2
In this session the first two examples still refer to design where
6 treatments are applied to 36 units. However, more complex
structures are placed on the units. In design ED2.1 a Latin
Square design is employed and in design ED2.2 a split-plot design.
In the third example design ED2.3 the number of treatments is
increased to 12. This has been done through the introduction of a
third treatment factor C with 2 levels, i.e. the treatments have
a 3 x 2 x 2 factorial structure.
**** ED2.1 Latin Square Design (Plan 2.1) ****
In this design two forms of blocks are imposed on the treatments at
at 'right angles' to each other e.g. rows and columns; animals and
periods. The intersection of a row and column defines a unit.
Define the treatment factors and read the basic data.
"
unit[nvalue=36] unitno
factor[levels=6] T
factor[levels=3] A
factor[levels=2] B
open 'sess2.dat1'; channel=2; filetype=input
read [print=*; channel=2] unitno, T, A, B, TE, PE
"
Define the blocking factors.
"
factor [levels=6; values=(1...6)6] row
factor [levels=6; values=6(1...6)] column
"
!!! The next stage is non-standard - apply randomisation and create
the variate for analysis.
"
random [blocks=row*column; seed=12345] T, A, B, TE
calculate [print=summ] V3 = TE + PE
"
End of non-standard stage !!!
"
"
Print data
"
print unitno,row,column,T,A,B,V3; fieldw=7,5,7,3(5),8; decim=6(0),3
"
Perform analysis of variance and keep residuals. Here a residual is
determined by the difference between an observation and the value
expected from the particular combination of treatment, row and column.
"
blocks row*column
treatment A*B
anova V3; residuals=RV3
"
Plot residuals against unit number
"
graph [nrows=20; xint=y] RV3; unitno
close 2
open 'sess2.dat2'; channel=2; filetype=input
read [print=*; channel=2] unitno,T,A,B,TE,PE
"
**** ED2.2 Split-plot Design (Plan 2.2) ****
In this design two types of block are imposed on the units. Six
large blocks are divided into three smaller blocks or 'main-plots'
each of which contain two units or 'sub-plots'. The levels of
factor A are identified with the main-plots and the levels of
factor B with the sub-plots.
"
factor [levels=6; values=6(1...6)] block
factor [levels=3; values=2(1...3)6] mainplot
factor [levels=2; values=(1,2)18] subplot
"
!!! The next stage is non-standard - apply randomisation and create
variate for analysis.
"
random [blocks=block/mainplot/subplot; seed=12345] T,A,B,TE
calculate [print=summ] V4 = TE + PE
"
End of non-standard stage !!!
Print data
"
print unitno,block,mainplot,subplot,T,A,B,V4; field=4(9),3(5),8; decim=7(0),3
"
Perform analysis of variance
"
blocks block/mainplot/subplot
treatments A*B
anova V4
close 2
delete block,mainplot,subplot,T,A,B
"
**** ED2.3 Randomised Block for 12 treatments (Plan 2.3) ****
In this design blocks of twelve plots are superimposed on the units.
Each block contains each of the twelve treatments once. The twelve
treatments are formed from all combinations of three factors A, B
and C which have 3, 2 and 2 levels respectively.
Input treatment factors and basic data.
"
factor[levels=12] T
factor[levels= 3] A
factor[levels= 2] B,C
open 'sess2.dat3'; channel=2; filetype=input
read [print =*; channel=2] unitno,T,A,B,C,TE,PE
"
Define the blocking factors.
"
factor [levels= 3; value=12(1...3)] block
factor [levels=12; value=(1...12)3] plot
"
!!! The next stage is non-standard - apply randomisation and create
variate for analysis.
"
random [block/plot; seed=12345] T,A,B,C,TE
calculate [print=summ] V5 = TE + PE
"
End of non-standard stage !!!
Print data
"
print unitno,block,plot,T,A,B,C,V5; field=3(7),3(5),8; decim=6(0),3
"
Perform analysis of variance
"
blocks block/plot
treatment T
anova V5
"
Repeat analysis using factorial treatment structure and keeping
residuals.
"
treatment A*B*C
anova V5; residuals=RV5
"
Plot residuals against unitno
"
graph [nrows=20; xint=y] RV5; unitno
"
Exercises Session 2
Exercise 2.1 The twelve treatments of example ED2.3 are to be applied
to the thirty-six units. However, the structure imposed
on the units is that 3 blocks of 12 units are sub-
divided into 6 main plots of 2 units or sub-plots.
File 'sess2.dat4' contains the factors T, A, B, C
and variates TE, PE. Based on examples ED2.2 and ED2.3,
generate the data for a split-plot design in which
factors A, B are identified with main plots and C with
sub-plots. Analyse the data.
"
stop