Glossary

 

% 

Is the prefix associated with constants.

# 

Is the prefix associated with functions.

address 

An address consists of one of the following.

1. @label(c)
2. @label
3. (c)

where ``label'' is a label, and c some valid channel number, depending upon context. If the channel number (c) is missing, then all current input files are searched for the label. If the label part is missing, the beginning of the file linked to the channel number is assumed.

assignment 

An assignment is a linear combination of elements, components, and scalars. Each assignment is known by an identifier.

base 

Bases are the fundamental organisations of elements into collections. In [B/D] they are identified by a sequence of alphanumeric characters, beginning with an alphabetic character. For most applications, single elements may be thought of as bases containing only that element.

belief store 

Every element has variances and covariances associated with it. There are a number (usually at least two) of alternative storage areas for these variances and covariances, corresponding to alternative variance-covariance specifications. Each such area is called a belief store. Belief store number one corresponds to the first belief store, and so forth. The number of belief stores may be found by issuing the LOOK:  command using the program  argument. See also expectation store.

case 

[B/D] ignores the case of all characters entered unless the characters constitute text to be used for titling, or to be printed out directly.

command history 

A record of the most recent commands issued from the keyboard is retained in the command history . These commands can be replayed using the !:  command, or deleted from the buffer using the /:  command. The command history can also be saved to a file using the KEEP:  command.

command line 

A command line is generally a [B/D] command, followed by apparopriate arguments to the command. The arguments must be separated from the command by at least one space, or by one colon. The command itself must consist of contiguous characters. Thereafter, as many spaces as desired can be strewn amongst the characters forming the arguments.

component 

Components are notional elements. They carry no data and have beliefs specified for them only functionally. They take no part in adjustments, but are useful as intermediaries in generating true elements.

constant 

A constant is a scalar defined by the C:  command. Its value is determined at the point where the C:  command is used, and remains unchanged until redefined.

current adjustment 

The current overall adjustment of one collection by all the fitted collections is termed the current adjustment. The current adjustment is undefined if there is no current fit. If there is a current adjustment, there may also be what we term a previous adjustment, and in this case there will be available results concerning the difference between the two.

data-carrier 

There are two kinds of data carrier: data-elements (which are not associated with any other kind of belief specification), and elements (which are).

data-element 

Data carriers have a status similar to elements, except that there are no beliefs of any kind associated with them, and they always carry some observations or scalars.

definition part 

Given a particular command, we call the part to the left hand side of the equals symbol the naming part, and the part to the right hand side the definition part of the construction. Thus here is the naming part, and is the definition part of the construction.

element 

Elements are the fundamental belief-carrying quantities, corresponding to random variables in traditional statistical analyses. In [B/D] they are identified by a sequence of alphanumeric characters, beginning with an alphabetic character. One or more period symbols might be used to associate indices with the element. Elements typically are associated with expectations, variances, covariances with other elements and so forth. Additionally, one or more observations may become available on some elements.

equation 

A legal equation is an infix sequence of operators and operands, such as . In addition to the usual operators described in §21.9 and §21.2 a considerable number of operators and operands peculiar to [B/D] are available, allowing access to [B/D] outputs. These are described in detail in the sections of chapter 21. A pair of round brackets is generally used to delimit an equation. In certain situations the delimiting pair of brackets, is redundant, and may be omitted. However, in doubt you should enclose an equation within round brackets - they cannot harm.

expectation store 

Every element has at least one expectation associated with it. There are a number (usually at least two) of alternative storage areas for these expectations, corresponding to alternative specifications. Each such area is called an expectation store. Expectation store number one corresponds to the first expectation store, and so forth. The number of expectation stores is the same as the number of belief stores, and may be found by issuing the LOOK:  command using the program  argument. See also belief store. The main commands which require expectations take them from the current default expectation store. This is initially store number one, but may be changed using the e  argument to the CONTROL:  command.

function 

A function is a scalar function defined by the F:  command. Its value is not determined at the point where the F:  command is used. Instead, the functional form is stored and evaluated therafter when called. Depending upon the arguments to the function at at a given time, the evaluation may change.

history file 

A file to which the command history  is being saved. This file is only generated if the history  argument to the KEEP:  command has been used.

identifier 

Identifiers are the names used for the various quantities in [B/D]. Each identifier is a sequence of alphanumeric characters, beginning with an alphabetic character. The maximum length depends upon the machine implementation. Some identifiers may have as a suffix an index part, either known or unknown.

index 

Each index is either a single alphabetic character or an integer, each preceded by the period symbol ``.''. Integer indices are known and fixed. Character indices are allowed to vary (to take several different possible integer values), and are intended to become known at some future time.

index list 

The names of elements, functions, and so forth, may include one or more indices. The index parts of an element, function, etc., name always come after the main name. When defining quantities which include indices, you may not define a name which contains both known and unknown indices. However, an accompanying equation may.

label 

A label is an alphanumeric string starting with an alphabetic character, and prefixed with the @ symbol. [B/D] command lines can consist of such labels, possibly accompanied by commentative text, separated by one or more spaces from the label. You may define duplicate labels provided that they are on different input channels. If a duplicate label is found on the same file, the former is overwritten by the latter. The labels on a file may be viewed, with or without commentary, by using the LOOK:  command with l  and lall  arguments.

log file

A log of the outputs produced during a session may be retained in a log file.

multiple lines

These consist of distinct command lines separated by the vertical line symbol |.

naming part 

Given a particular command, we call the part to the left hand side of the equals symbol the naming part, and the part to the right hand side the definition part of the construction. Thus here is the naming part, and is the definition part of the construction.

ordering convention 

The following convention is adopted for the ordering of names. The initial part of the name, up to and excluding any digit, is considered first, and the basic ordering of the two names depends upon these parts. Thus `x10' is considered to be `left' of `xa2' as `x' is `left' of `xa'. Wherever the alphabetic parts are identical, the overall length is next taken into consideration, with shorter overall names to the left of longer names. Thus, `x2' is to the left of `x10'. Finally, when both the lengths and the alphabetic parts are the same, a straight alphabetical comparison is made between the two. Thus `x11' is to the left of `x12'.

previous adjustment 

The previous adjustment is only defined when two successive related adjustments (typically the first a simple adjustment and the second a partial adjustment) have taken place. In this case there will be available results concerning the difference between the two, representing the partial adjustment.

rounding 

For the purpose of conditional action, for example during use of the IF:  command, we represent ``true'' by any real number whose value rounded to the nearest integer is unity, and ``false'' otherwise. A number which is exactly halfway between two whole numbers is rounded to the whole number with the greatest magnitude.

search path 

A search path tells [B/D] where it can find particular external files. Every search path typically includes the current directory as the first directory searched for a file. For macro input files, the default search path is set by the environment variable BDMACROS .