String assignment

!Assign

An !ASSIGN name string qualifier has been added to extend coding options. It is a high level qualifier command which may appear anywhere in the job, on a line by itself.

The syntax is, beginning in position 1,
!ASSIGN name string
and the defined string is substituted into the job where $name appears.
string is the rest of the line and may include blanks. If !< !> encloses string, string may extend over several lines, which are concatenated.

For example

 !ASSIGN TRT xfa(Treat,1)
 ...
 ... $TRT.geno
 ...
 $TRT.geno 2
 $TRT 0 XFA1
 ...
 geno


 USmatrix !<  20
   10 30
   10 10 20 !>
 ...
     us(Trait,$USmatrix).Geno

Restrictions

A maximum of 50 assign strings may be defined.

The combined length of all strings is 1000 characters.

name may consist of 1--8 characters but should not begin with a number (see command line arguments).

Dollar substitution occurs before most other high level actions. Consequently, ASSIGN strings and commandline arguments may substitute into a !CYCLE line.

FOR loop

The !FOR ... !DO ... command is intended to simplify coding when a series of similar lines are required in the command file which differ in a single argument. The list of arguments is placed after !FOR and the command is written after !DO with $S indicating where the argument is to be inserted. list may be an ASSIGN string since they are processed before the !FOR statement is expanded. Furthermore, if list is entirely integer numbers, i:j notation can be used. For example

 !ASSIGN Markers M35 M75 M125
 !FOR \$Markers !DO !MBF mbf(Geno,1) markers.csv !key 1 !RFIELD\$S !rename \$S
 ...
 ... ~ ... !r \$Markers

The aim here is to generate the 3 !MBF statements required to extract markers 35, 75 and 125 from the marker file markers.csv. The names of model terms must begin with a letter, hence the marker names are the letter M followed by the position number. Adjoining !RFIELD and $S results in the M in the marker variable name being ignored when parsing the MBF statement so that only the number is picked up by the !RFIELD qualifier. Now, to add another marker to the list, just add it to the ASSIGN statement.

Restriction: list and command are both limited to 200 characters.

Conditional assign

!IF test text
One form of the IF statement is !IF $1 == wwt !ASSIGN M1 brt DamAge
which makes this !ASSIGN statement active if $1 is 'wwt'. This has been used when performing a large number of bivariate analyses with trait specific fixed effects being fitted. So

 ...
 !IF \$1 == wwt !ASSIGN M1 brt DamAge
 !IF \$1 == ywt !ASSIGN M2 brt DamAge
 ...
 \$1 \$2 ~ Trait at(Trait,1).(\$M1 )  at(Trait,2).(\$M2 ) ,
  us(Trait).animal
 ...

CALC statements

!IF condition !CALC \em expression}
The purpose of these statements is to allow users to specify weights and response variables that change each iteration. .
The condition EVEN|ODD|EVERY controls whether expression evaluated at EVERY iteration (except first), every EVEN iteration, or every ODD iteration (except first).
expression is an algebraic expression:
The following variables are internally defined.
R1 ( R2) residual (from .yht) from last iteration for trait 1 (2)
H1 ( H2) hat value (from .yht) from last iteration for trait 1 (2)
W1 ( W2) weight variable from last iteration for trait 1 (2)
Y1 ( Y2) response variable fitted in last iteration for trait 1 (2)
Gi is the ith gamma parameter value (unscaled by σ²) obtained from the last iteration
Li is the ith variance parameter as used in fitting last iteration
Pi is the ith variance parameter value obtained from the last iteration
Vi is the ith data variable

Note that W1, W2, Y1, Y2 may be redefined.

The permitted operations are +, -, *, / and \.

Permitted functions are ABS(), EXP(), LOG(), SQRT().

Multiple expressions may be listed in one statement, separated by ';'.

ASReml can parse complex expressions in CALC statements (unlike the case of transformations).

Specify the !DEBUG qualifier on the command line to get ASReml to print intermediate results to check what is happening.

Example

HGLM algorithm by Felleki and Ronnegard
  Yval Xterm * Zterm *
  Gval !=Yval !-.6   !*V4
  Ywt !=1. Gwt !=1.
   !MAXIT 100   !ASUV   !EXTRA 4
 !IF ODD !CALC W1=EXP(R2-Y2)

 !IF EVEN !CALC  S1=1/W1; T1=H1/S1;  Z2= R1*R1/(1-T1)
 !IF EVEN !CALC Y2=LOG(S1) + Z2/S1 -1
 !IF EVEN !CALC W2 = (1-T1)/2

 Yval !WT Ywt Gval !WT Gwt ~ Trait Trait.X,
 !R us(Trait).Z
 residual units.diag(Trait)

This snippet of code is from the job used by Felleki etal. (2012) to fit a bivariate double hierarchical generalized linear models.

Restrictions

This option is developmental and should be used with special care.

Return to index