(1) TED has no longer a default value, i.e. you are now requested to specify what item you wish to edit. The previous versions defaulted to NOTEBOOK.
(2) CLED the command line editor now automatically executes the corrected command, when leaving CLED by typing a blank line, i.e. CLED now does implicitly a :X.
(3) The LAG() function now has two arguments (instead of one): the first argument is the variable reference, the second then number of lags (it might be negative).
(4) User defined maps now have a different format, therefore old eda files with maps (e.g. the second WA of the old test file) will no longer load correctly (error message). As this has been a very incompletely documented feature, probably no one has ever used it.
(1) This level introduces the Zvariables (result variables) everywhere where other letter variables A..Y may be specified. Zvars have two letter names starting with a Z. As in many places you may specify letter variables AND case identifier the same problem as with single letter case identifications arises now with two letter case identifiers starting with the letter Z (upper and lower case). For this reason the same rule applies as for single letter variables: enclose the string into quotes (you may omit the ending quote) in order to cause evaluation as case identifier. For example LIST CASE=ZH which caused the case ZH to be displayed will now search for the ZVAR H, therefore you need to specify your command as LIST CASE='ZH'.
(2) TED: In earlier versions the line editor (editing the current text line) was invoked by the : symbol. In these versions it was not possible to edit TED command lines as in the normal EDA mode. This level now offers this possibility. As a consequence the text line editor symbol has been changed to ",".
(3) The CANON command has been altered and now has a different syntax. In particular in default mode the command no longer queries for the variable list (specification of a double variable list on the normal command line). In fact this solves a problem with the previous command, which did not permit inclusion of variables which where not contiguous and not starting at the top of the WA.
The SCORES option is no longer implemented as sub-command, but as a ordinary option. As a consequence the NEW parameter is no longer needed.
(4) *WRITE for raw data output (including SPSS, TRANSPOSE) now generates its format automatically from the variable min/max and the currently set decimals. Therefore the *WRITE command no longer asks the user to specify the format. Command streams having a format specified should be modified: (a) either the format is removed (b) or the USERFORMAT parameter is added; then the command will work as before.
The SPSS option now generates SPSSX set-ups. To obtain SPSS setups as previously add the OLD option.
(5) The CFIX command has been reworked. The input configuration is now stored in C1 and no longer specified as variables.
(6) The C1/C2 commands have been enhanced. The only option which does not work as before is the C1/C2 LOAD, which now works like most other copy options. If nothing is specified the configuration is copied into free locations in the WA, searching from the beginning.
The C1/C2 LIST command now blanks coordinates which are smaller (absolute value) than a certain threshold. The default is 0.05. Therefore some values that appeared in previous versions are now shown as blanks.
The DESCRIBE and DS displays look different (nicer, I hope).
ALLVARS mode: ALLVARS mode includes all variables only if no vlist is present on the command, say FACTOR. If a vlist is specified on a command it behaves like in ALLVARS OFF mode, i.e. includes only the variables specified.
OUT has slightly changed options in the parameter field. Especially there is now a PFIL option to put the result to the print file.
DOC (1) The DOC FILE command has been changed into DOC CREATE; DOC FILE means to write the output of DOC commands to the RAWOUT file (FILE is used to achieve consistency across commands). (2) The displayed document may no longer be preceded by a string, specified with a "text". This option has been suppressed, because not very useful and confusing. If that functionality is needed, copy the display to a file using the new FILE option and then use the toolbox to modify Toolbox: PRINT is no longer a synonym of TYPE, but is now a command printing a file on the line printer.
PROFILE: The older version used "0" "1" ... symbols to show case or group position within a distribution, i.e. the first case/group was labelled "0"; this seems rather illogical when using 0 for the 1st group etc. therefore "1" is now the first "0" being the 10th... (makes more sense, hopefully).
Built in case ids and maps have been removed from EDA as announced. Therefore the CASID CH, reps CASID EURO commands do no longer exist. However these maps/casids are available as map/casid archive. Therefore if this archive is installed at your site you just have to say CASID DEFINE CH resp. EURO to do exactly the same as CASID CH/EURO. The corresponding MAP will then be associated with the MAP command.
HISTOGRAM : If a selection is active EDA no longer uses min/max from all cases, but computes them on the selected cases. The GLOBAL option has been added in order to use the global min/max instead.
BOXPLOT PARALLEL has also a GLOBAL option, in order to force scaling on the global min/max with selections. If you wish to get the same output as previously you need to specify GLOBAL.
PLOT RECOMPUTE: To make scaling consistent with other commands EDA now rescales automatically with cases selections; as with HIST and BOXPLOT parallel you will have to specify GLOBAL in order to obtain global scaling.
ELABEL and EDESC (EDIT commands) no longer exist; they have been replaced with a slightly different syntax with the LABEL command in normal EDA mode. (Expressions, undocumented "feature") Unconditional expressions, i.e. LET/CALC commands are no longer selection sensitive. In previous versions only the selected portions of variables were modified. This caused sometimes problems with non yet existing variables and always with MATRIX, C1 and C2 references.
From this version on all expressions no longer interact with the selection feature; an IF no longer turns the current selection OFF, nor is it altered (unless of course INCLUDE/ EXCLUDE or DO are used).
HISTOGRAM: HISTOGRAM GVAR is now used to display a current GVAR and not to store it; this is now done using the HISTO STORE_GVAR option.
DICHOTOMIZE: Now there is a default limit of 10 dummies to be generated from a single variable (MAXCAT); in previous version default was no upper limit.
VHIERARCHY FLEXIBLE: earlier versions queried for the additional parameters; this was rather inconvenient when playing with various values using macros; they now have to be specified with a new Parameter=(list); absence of Param= produces an error, i.e. the old style input is no longer supported.
CONFIG the SET option is now called SET=(ndim,ncoord), i.e. the vlist field is no longer used to indicate the size of the config to be filled. [The old form lead to confusion]
With this version the workfile commands (*ATTACH, *SAVE, *LOAD etc) no longer exist as separate commands. They are now found within a single command *WORKFILE as options. These commands will need installation, i.e. they are normally no longer available, except when installed at EDA system generation. Anyway use of this commands is not recommended for new users.
MINISS and MDSCAL have no longer a NODISPLAY option. Use the global switch /d instead.
The manual has now a different structure, i.e. the commands are no longer in alphabetical order, but in functional groups. LIST: the Length= has been replaced by Width= parameter, because the SHOW and LIST command now use the same display routines; this avoids some annoying inconsistencies. If you use LENGTH with the normal LIST command, a message will suggest that you use width instead.
STANDARDIZE to archive some consistency in the exploratory logic of EDA, the default is no longer the classical standardization (means/standard deviation) but the exploratory (median and midspread).
ZVARS $1 instead of Z1 Pseudofunctions )NVAR replaced by $NVAR value substitution $GE replaced by $DAT.var#.GEZVARS: even if they are still called ZVARS (or result variables) they are no longer used with a Z in front of them (except Z$, i.e. the string ZVAR). You now will have to write $0, $1 etc. instead of Z0, Z1 etc. This also means that casids starting with a Z, e.g. ZH need no longer be quoted to produce appropriate substitution (this was the main reason for the change: achieve a minimum of consistency.
pseudo-functions: the ) reference in normal EDA syntax, as well as the function reference (e.g. NVAR() ) has been replaced by a reference starting with a $ in all cases, i.e. $NVAR. Some of these references may have arguments: $NOC.1 gives the number o cases in variable number 1; $DATA.XV.1 returns the value of the first case of the variable 'XV' (this replaces the old CASE=$1 substitution mechanism, where the variable were taken from the first variable in the vlist).
value substitution as in CASE=$GE. See ZVARS for details.
SCAN (EDIT command) no longer exists, all options have been included into the CHECK command (normal EDA mode). The only difference with the options available with SCAN (there are new options) is the fact that you may specify a variable list for variable related checks. With EDIT this was not possible due to the structure of the EDIT module.
The /d and /p switches were not available for all commands; this is no longer true. All commands in normal EDA mode now accept the /d and /p switches for controlling the display and print file output. (Currently TED and the toolbox are not completely adapted to this).
The *READ DATA command no longer exists, as all its facilities have been incorporated into the *READ RAW command. Files for *READ DATA cannot be read with *READ RAW without conversion. There is a conversion tool available. This also means that the BUILD command no longer works as it did previously. Old input files to BUILD require minor modifications in order to read them directly with *READ RAW.
TRACES is now much more powerful, the TRACES, NOHINGES and EIGHTS options no longer exist: TRACES has a sligthly different meaning (special case of TRACE=); NOHINGES is replaced by MEDIAN; EIGHTS is replaced by TRACES=.
BREAK and XTAB: The threedimensional display shows the information in a different order (user demanded modification).
For this reason this section lists commands or program areas which are considered deprecated, i.e. features which in a near future will be completely altered or completely removed (unless the user community wishes to maintain such a feature).
With the introduction of the WA directories (and the PUT/GET family of commands) the old style *READ EDA and *WRITE EDA commands have become obsolete, as well as system files storing more than one work area. The *READ/*WRITE EDA commands will be dropped in version 2.4 of EDA. If you still use these commands refer to the section on files and namely the PUT/GET and PUT/GET EXTERNAL commands. [In fact PUT/GET EXTERNAL are synonyms for *READ/*WRITE EDA, use these alternative command forms instead so you will not be surprised when the *READ EDA/*WRITE EDA commands disappear.
The possibility of writing several work areas to the same file is not longer really useful when you have WA-libraries available. In addition to that their use was not really unproblematic, as deleting the file meant deleting all the work areas within quite a few people ran into trouble... Their use is no longer recommended. Support for these kind of files will be dropped in future versions. If you ever used this feature and continue to use it... you should read the manual to find out how much easier it is to use WA libraries.
Many options have been added for coding (various forms) in this release. Coding symbols have been systematized (there were some incoherent symbols used). Furthermore with distributional coding there will - in many instances - more symbols: Until now we used to code (often) far, out, adjacent and within, without any distinction of low and high fars, outs and adjacents. Now this is - where useful - always done.
Various options dealing with coded LISTs have been added. If you look at the options of the LIST command you will notice that some have change their name (more systematic vocabulary). Older options names are still supported, but they no longer appear in the documentation.
Note also that some displays might look a bit different, as some values will be coded in the next lower or higher bin, due to rounding. (This will only appear with data having many identical values).
If you don't like the new codes use the SET GRAPH command put back the old symbols use the SET GRAH command (for a permanent change you can also modify your profile). If you do not want to distinguish low/high values you can use the SET GRAPH DIST SIMPLE command.
HIERARCHY/VHIERARCHY: hierarchical trees look slightly different if many objects are merged on the same level (same as displayed, not as computed); then the links are mostly omitted, except for the first and the last object (lowest and highest). This makes the tree (hopefully) more readable. Still there are limits to what can be done on a character based screen, i.e. it is always possible that - especially on lower levels - that you are not able to decide to what cluster an object or a group object belongs, i.e. you will need help from the additional tools provided with the TREE (sub)command. If you don't like the new way of showing the tree you will have to add the FULL option when producing one.
MAP DISTRIBUTION is now called CODE DISTR and no longer produces symbols of the form out, -far, but uses now the standard coding symbols used in the whole EDA System.
SHOW CODED now uses the standard coding symbols. Furthermore the
LE= GE= etc options have been replaced by IF
For experts: center/min/max appear at the end of each variable (last
three values). The record type is set to 8 (therefore older versions
just skip over these records). If (for hackers) you got this kind
of files and you need to read them on older version you can edit them
by setting the record type to 0, and removing the last three values
from each variable (if these three values are together with other
values on the same record (line), you can leave them as they are, as
EDA will just ignored them.
Variable lists:
due to the introduction to vlist editing (+ or - sign to signal that
you want to modify the current list by removing or adding a new list)
a variable list like BOXPLOT -5 is no longer interpreted the same way.
Before version 2.2 this meant that you wanted to produce a boxplot for
variables 1 through 5, i.e. the omission of a starting variable
meant variable 1. It now means that you want to remove variable 5
from the current list.
DS, DESCRIBE, VARS commands: Due to the introduction of new options the
INTO option has been changed into AS.
STEMLEAF now creates by default LO and HI stems, whenever outliers are
found in the data. If you want to produce stem and leafs the old way,
you will have to specify the NOHILO option.
QUIT, STOP, PRINT, LAST QUICKPRINT and PRINT REVIEW:
These commands had a
DIAGNOSTIC: no default values accepted: An option is now required.
ROLL function
The ROLL function was in fact misnamed, i.e. it drew a sample WITHOUT
replacement. It is now correctly named DRAW and there is a NEW roll()
function drawing a sample with replacement.
QUIT and PRINT KEEP (NOPRINT):
On some EDA implementations the KEEP option
(or K answer to the prompt when QUITtin) will ask you to enter a name
for the print file, i.e. instead of always creating a EDA defined
print file name you might use your own (if you hit the return key you
will get the same name as before).
DISTANCE, MDS, CLUSTER, VHIERARCHY, TSCALE, MINISSA, DIS() function:
A new SET POWER command is used to control the default distance
to be computed. The various options used to control the distance metric
are now used to override the default setting,
i.e. default is now global and
no longer local. Unless you change the global default value using
SET POWER, EDA behaves as in previous versions (default Euclidean
distances, local options override that default).
Due to the introduction of the SET PLOT options default sizes and
plot symbols now depend upon the global default settings.
E.g. PLOT 1 2 will only produce the same plot as in previous versions
if the defaults have not been changed.
.shead Changes for macro users
All lines starting with an exclamation mark (comment mark) are
treated as comments, i.e. the single comment mark no longer displays
the line.
As the documentation was not too clear about this point you did
probably not use this "feature". In all cases the > symbol provides (and
provided) a better solution for displaying messages to the user.
In previous versions QUIT and STOP caused EDA to terminate in all cases.
In macros and command files QUIT and STOP now
issue an error message. If you really want to stop EDA from within
a command file or macro you will need to add the EDA option, i.e.
QUIT EDA.
Single line macros (abreviations) have been altered completely.
Former versions had a FOR or INDEX option (the same as with EXECUTE
and the expressions). This option could enter in conflict with syntax
analysis (S= E= FOR INDEX options on the commands) and was never really
useful, as the same could be achived with the EXEC "command" for ...
command.
Furthermore there was a not very useful "s" to add options to
the macro.
Now the rules are simple: EDA analyses the macro-line and its options,
substitutes then the macro definition and analyses the new options.
This means that the macro-defined options always take precedence over
other options, however options not set in the macro definition can
be used freely to change the macro's behaviour.
Version 2.2
CANONICAL command: CANON vlist1&vlist2
You now should use the "&" to separate the sets, i.e. the ";"
separation character has be replaced by "&" (due to the introduction
of multiple commands on a single command line).