Features in earlier releases

Home Page	Overview
New Features	Downloads & Prices
P-STAT documentation	Technical Support
Contact P-STAT	Unistat for Windows

Features of version 2.22 release 10

Formatting changes:

Fractional numbers canot always be precisely represented even in 64 bit precision.

This can lead to a result of 14.5, computed one way, rounding to 14 and another 14.5, computed a different way, rounding to 15. This is due to slight differences in the 64 bit representation of the two 14.5s.

The P-STAT formatting routines have been changed, and will now cause such numbers to consistently round up.
Error messages are much improved when invalid arguments are given to PPL functions.
An error was introduced on October 14, 2004 in the /// character concatenate function; this is now fixed.

Features of release 9

The TURF command has 4 important enhancements.

The SIZE identifier can be used to specify multiple sizes. For example SIZE 4 TO 8 would cause those 5 sizes to be done.
FREQ scoring has been changed in two situations.

The first is when reach.threshold is not 1, or when response weights and/or item weights are used. The data from all cases, reached or not, had been used to compute the FREQ score. Now, only reached cases contribute to the FREQ score.

The second is when response weights and/or item weights are used. The freq score is now the SUM of the values rather than the COUNT of the non-zero values. This better reflects the impact of a given combination.
A new option, SET.MISS.TO.ZERO, causes missing values on the analysis items to be converted to zeros. This saves writing some PPL when this change is wanted.
It is now possible to terminate a run that appears to be taking too much time. Press the F2 function key and confirm that you indeed want to terminate the run. All the files are closed properly and and all completed sizes are reported in the output files. This permits a tidy way out if you realize that you have asked for a size combination that might take many hours/days.

TABFILE.IN enhancements

TABFILE.IN is now much less restrictive regarding the supplied names of variables. It will try to cobble a legal name from whatever it is given. If too messy, a var17 type of name is generated.
Extensive tests have been added to make sure that the values within a case end with the delimiter, and that the value ending the case ends with the end-of-record, which in most files is a CR-LF.

SURVEY changes

Concatenation of extended variable name strings is no longer limited to 2 strings per line.
A new subcommand QUESTION IN LABELS.AREA formats the question in the margin of the labels are. This is supported in regular, postscript and tabbed output but not in the transposed file format.
The TOTALS.AREA can now contain 2 entries. Thus both the total count and the base can be in the same section. In addition the percents are now printed when the base is in the TOTALS.AREA and the percents are based on the on the good count.
Several minor problems have been fixed. NEVER SKIP no longer leaves blank lines before a bottom title. NO LEAD BLANKS now applies to title lines.

General System Changes

The maximum length for P-STAT system file names has been increased on all systems to 16 characters. This does not include the 3 character extension. The P-STAT command ALLOWABLE.NAME can be used to decrease the size if necessary.

PC/Windows enhancements

The alternate characters on the number pad such as plus and minus are now accepted.
It is now possible to do more complex PPL using the menus including simple DO loops.

Features of release 8

TABFILE.IN and TABFILE.OUT now allow alternate delimiters instead of the default horizontal tab character. BLANK or COMMA (or indeed any ascii character) can be used.

Two identifiers, NUMERIC and CHARACTER, have been added to TABFILE.IN to specify the datatypes of selected variables.

Minor changes have also been made to the PC/Windows menus.

Features of release 7

The PRINT command has been rewritten. for PC/Windows, Linux and Unix. It is used to send either a text file or a P-STAT system file to the system printer or print queue. Since the P-STAT system file is a binary file that cannot be printed directly, a LIST command is automatically generated which writes the contents to a text file. The text file is given a generated name that will not be confused with any other file. A generated CLOSE and a generated PRINT with the text file as its argument completes the process. The PRINT command is fully documented in the P-STAT help file.

Version 2.22 release 5

For PC users there is a new front end on all systems. It uses familiar windows features such as the file manager, the print manager, pull down menus and message boxes. Those users who always run from scripts in batch mode will not notice any difference. Users who run interactively will find it easier to print and to edit commands.

The major feature of release 5 is one that you will only see if you make an error using one of P-STAT's functions. The error messages are greatly enhanced by including the function name in the error message.

November 12, 2003 is a minor upgrade. For PC/Windows the numeric keypad is again recognized. For all versions SURVEY has 2 new subcommands, USE RESPONSES and USE PERCENTS. The usage is documented in the online help file.

A new date/time function WEEK.WITHIN.YEAR has been added. Either an ABSOLUTE week or a CALENDAR week can be used, and the beginning day of the calendar week can be specified.

WEEK.WITHIN.YEAR ( date, integer ) >>> integer

This function returns the week number within the year for the supplied date. The range can be 0 to 53, depending on the date and on the calculation method. There are two methods for determining what constitutes week one of a given year.

The first method is simple: the first week goes from Jan 1 through Jan 7. This can be called an ABSOLUTE week.

The second method makes use of a calendar week, which is defined by ISO 8061 as going from Monday through Sunday. The first week is the first CALENDAR week that contains a sufficient number of days within the current year.

Sufficient can be set to 1 through 7; the ISO standard is 4. This function assumes a Mon-Sun calendar week; a different calendar week can be given in the function.

  ***********************************
  *  examples using ABSOLUTE weeks  *
  ***********************************
  
  Week.within.year ( 'jan  3 2004', 0 ) = 1
  Week.within.year ( 'jan  5 2004', 0 ) = 1
  
  Week.within.year ( 'jan  8 2004', 0 ) = 2
  Week.within.year ( 'dec 31 2004', 0 ) = 53
  
  ************************************
  *  examples using the default      *
  *  Monday to Sunday calendar week  *
  ************************************
  
  mon  tue  wed  thu  fri  sat  sun
				   1    2    3    4
	5    6    7    8    9   10   11
   12   13   14   15   16   17   18
  
  Week.within.year ( 'jan 5 2004', 1 ) = 2
  Week.within.year ( 'jan 3 2004', 4 ) = 1
  
  Week.within.year ( 'jan 5 2004', 4 ) = 2
  Week.within.year ( 'jan 5 2004', 4 ) = 2
  Week.within.year ( 'jan 5 2004', 7 ) = 1
  
  **************************************
  *  examples using an alternative     *
  *  Sunday to Saturday calendar week  *
  **************************************

  sun  mon  tue  wed  thu  fri  sat
						1    2    3
	4    5    6    7    8    9   10
   11   12   13   14   15   16   17
  
  Week.within.year ( 'jan 3 2004', 1, 'sun' ) = 1
  
  Week.within.year ( 'jan 3 2004', 4, 'sun' ) = 0
  Week.within.year ( 'jan 5 2004', 4, 'sun' ) = 1
  
  Week.within.year ( 'jan 1 2004', 7, 'sun' ) = 0
  Week.within.year ( 'jan 2 2004', 7, 'sun' ) = 0
  Week.within.year ( 'jan 3 2004', 7, 'sun' ) = 0
  Week.within.year ( 'jan 4 2004', 7, 'sun' ) = 1

The major new feature in release 4 is "EXPAND".

EXPAND is a PPL statement that projects the values of one or more input variables into a set of new variables, each associated with a specified value in the input variables.

In it's simplest usage, EXPAND uses one input variable to create a group of new zero/one variables. Each case begins with the new variables set to zero. Then, if the value on the input variable is one of the specified values, the associated new variable is set to one.

The input variable can be numeric or character. The output variables are always numeric. These new variables are sometimes called "dummy" variables.

Several input variables can be expanded together. The output variables can then be set to:

(1) one if ANY of the input variables has the associated value. This is the default.

(2) the NUMBER of input variables that have the associated value.

(3) a one if the first input variable has the associated value, otherwise a two if the second input variable has the value, and so on. In other words, the RANK of the value.

An Example of EXPAND

Suppose variable BREAD is coded 1 through 4, with 1 meaning rye, 2 meaning wheat, 3 meaning raisin and 4 meaning white. The PPL statement

[ expand bread, values 1:4, gen rye wheat raisin white ]

will generate four new variables named RYE, WHEAT, RAISIN and WHITE. If a given case has a 1 on BREAD, the value for RYE for that case will be set to one and the other three to zero. If BREAD is two, the second new variable, WHEAT, is set to one and the rest to zero, and so forth.

The new variables are placed after the last current variable. In the VALUES phrase, either 1:4 or 1 TO 4 could have been used, they mean the same thing.
TURF: Enhancements to the TURF command includes a new option allows you to try only the combinations that satisfy logical constraints which are specified in the command. The output files will contain the results for the combinations that were actually evaluated. This saves having to search the REACH output file for the combinations that make sense.

You might specify that any combination to be used must have at least 2 but no more than 4 of a specified group of variables. This is done with the USE identifier. For example:
```
       TURF xxx, size 8,
                 use   aaa         bbb to ddd   1 1,
                 use   eee to ggg  jjj to mmm   2 4,
                 use   yyy         zzz          0 1 $
```
In the above command, the only combinations that will be evaluated are those that have one and only one variable from the first group, and at least two but no more than four variables from the second group, and no more than one variable from the third group.
In SURVEY some problems with the printing of significance tests have been fixed. The FPC, finite population correction, has also been corrected to work with f tests.

Version 2.22 release 1 & 2

This release consolidates all the changes in version 2.21 releases 1-10 and contains many new features.

BUILD: an option, CASE.RECORDS, has been added to free format BUILD to process external files in which carriage returns and line feeds occur as other than record delimiters. The end.of.case character is used as the record delimiter. CR and LF characters can be dropped, kept, blanked or changed. Use HELP BUILD.FREE $ for details.
TABFILE.OUT: if a horizontal tab character is found in a character value, a blank is written instead to the output file. The same is done for carriage return and line feed characters.
DATE and TIME: 2 additional PPL functions have been added. EXTRACT.WEEKDAY returns the name of the weekday. DAY.WITHIN.WEEK returns 1 to 7, default is for 1 to mean Monday.
COUNTS: suppose a variable has 2 values that are slightly different but nonetheless produce the same result when formatted. For example, the constant .3 and the sum of .1 and .2 look like they should be precisely the same. However they may not be, due to rounding of binary approximations of decimal numbers which causes a difference in the final (64th) bit.

Running these PPL commands demonstrates the problem:
```
if .3 ne .1 + .2, put 'different' $


if .5 ne .2 + .3, put 'different' $
```
The first, on both SUN and PC, will say 'different', the second will not.

Given two faintly different values, COUNTS has treated them as separate values, even though the formatted appearance of the values as shown in the counts output was the same. In other words, .3 would appear twice in the output file. The LIST command, which prints the counts output, would then blank the second .3, to make the output less cluttered.

The COUNTS code now combines values that have the same formatted appearance.
There are several enhancements to the SURVEY command.

The tests of proportion assume that the distribution approximates the normal curve. In the past we have left the burden of determining normality in the tests of proportions to our users. This is now changed.

When tests of proportions are requested, a test is done to see if the smaller of Np (the number expected) and Nq (where q is 1-p) is at least 10. We have also changed the test value when tests of independence are requested from the current setting of 5 to the more conservative value of 10.

There are also minor enhancements to the RANK subcommand, in the tables of contents, and in page controls.
- CEIL is a new function that takes a numeric input and produces the smallest integer that is greater than or equal to the input value. Thus,
  
  CEIL (-4.7) = -4
  CEIL ( 2 ) = 2
  CEIL ( 2.1) = 3
- FLOOR is a new function that takes a numeric input and produces the largest integer that is less than or equal to the input value. Thus,
  
  FLOOR(-4.1) = -5
  FLOOR( 2 ) = 2
  FLOOR( 2.9) = 2
- CLAG is a new function that performs a lag on a character argument, which can be an expression.
  
  GEN PREVIOUS.TITLE:c30 = CLAG( JOB.TITLE, 12 )
  
  This would take the JOB.TITLE value from 12 cases ago and copy it into PREVIOUS.TITLE of the current case. The second argument, the lag depth, must be an integer constant from 1 to 500.
- LAG and DIF: the first argument had been restricted to variable names only, expressions are now allowed.
- DATE and TIME: about 40 PPL functions have been added. These accept date or date/time input and do things like add days to a date, or determine the number of hours between two date/time values.
  
  These functions accept dates in character form like
  
  'Mon June 17, 2002 12:30:12.123' .
  
  The time field, which allows milliseconds, can be omitted. The year can be from 1753 to 2999.
  
  Six logical operators (DATE.LT through DATE.GT) have been added. These are used to see if one date comes before or after another. For example
  
  [if birthday DATE.GT 'jan 1 1950', delete]
  
  You can, for example:
  
  add days and hours to a date and get a new date.
  get the difference in hours between two dates.
  change the year in a date value.
  determine if one date precedes another.
  get the fiscal year or quarter of a date, and
  extract the century (the 19 in 1997) from a date.
  
  For sorting, there are functions that take a date and return seconds or days since jan 1,1753, or seconds since midnight.
  
  The default is to use English weekday and month names. DATE.LANGUAGE GERMAN $ is a new command that causes German names to be used instead.

Version 2.21 release 10

Release 10 has minor enhancements and bug fixes.

The PPL PLACES function has been changed so that a value like 0.35, when one decimal place is requested, will routinely round up. This is done by adding an extremely small additional increment when doing the round. This change has also been made to the PPL ROUND function.
A problem in the BALANCE command has been fixed. The happened when the number of control variables and categories required the use of a second index block.
Several SURVEY problems caused by sparse data have been fixed. These are described in the help file.

New features in release 9

INQUIRE xxx$ This is a new command which determines if a file (in this example, xxx) is there or not. A system variable (.inquire.) is set to 1 if it is there and 0 if not. The file is not read by this command. A one-line report is written, which can be suppressed by using 'verbosity 0' in the inquire command.
WILDCARDS have been limited to forms like abc? (a prefix) and ?abc (a suffix). They now can have a number of question marks, like ?ab?c?de?. A test has been added to prevent a variable from being inadvertantly referenced twice in lists using wildcards.
In the SURVEY command, new features of the signficance test subcommand, PAIRED.VALUES and PAIRED.VARIABLES, had problems when changing the test value from .95 . These have been fixed.

New features in release 10

TURF has additional weighting options. ITEM.WEIGHTS can be used to reflect the cost of items, and RESPONSE.WEIGHTS can be used to reflect the intensity of the responses.
TURF.SCORES is a new command that writes each case's reach score on a specified combinations of items to an output file.
TABFILE.IN is a new command that reads a tab-delimited ascii file and creates the corresponding p-stat system file. For example, this can read a tab-delimited .txt file made by EXCEL.
TABFILE.OUT is a new command that reads a p-stat system file and creates the corresponding ascii tab-delimited file. This could then be read by EXCEL.
Two changes in printing have been made.

(1) If a file used for printout is reopened, its attributes when it was closed are re-applied.
(2) The default formfeed character on PC windows versions has been changed from a "1" to an ascii 12.
The AMONG PPL logical operator has been enhanced. It now allows .M1. and .M2. and .M3. (and .M. for all three) to appear as test values. For example:

if case.number among ( 102 to 255, .m1. ), etc.

When case.number is missing 1, the above result is true. When case.number is missing 2 or 3, the result, as before, is missing.
SHOWBYTES now has a FIND option. This allows the file to be searched for a specific byte or set of bytes. When found, a display occurs showing the bytes before and after the matched area. Another search is then attempted, and so on.
SURVEY enhancements include NO LEAD.BLANKS, a new identifier, several minor enhancements, and the following new subcommands:

RANK SUMMARY
REPORT ALL RESULTS
SVAR.LABELS
BVAR.LABELS
The SPSS.OUT command has a change to improve the precision with which very small numbers are converted to base 30 notation.

New features in release 7

User-defined arrays have been added to the PPL language. These arrays can have up to 7 dimensions, and can be numeric or character. The rules for using these arrays are described in the P-STAT help file.
Standalone PPL now allows nested DO loops which are especially useful in initializing arrays.
DBASE.OUT has a new option to specify the year byte in the header date of the new file. This overcomes a problem with packages that do not support a year byte of 00.
There are three improvements for PC/Windows users.
Directory and file names with blanks in the name are now allowed.
The windows behavior has been improved so that control returns to the proper line after a screen save or a window minimize. This is especially useful in DATA.ENTRY.
The Z option for full screen editing in the P-STAT editor now handles character delete in an intelligent way.

The major feature of release 6 was support of TURF , Total Unduplicated Reach and Frequency, analysis.

Release 5 provided some enhancements to the SURVEY command. For full details see: SURVEY enhancements in P-STAT 2.21 Releases 5/6 (May 29, 2000)

P-STAT macros for the social scientist

If you are interested in social network analysis, a series of of P-STAT macros is now available on the statlib server. Click on P-stat and look for netstat1.tar.gz.

Go to Statlib repository for P-STAT macros

New Features in P-STAT 2.21 release 3.

There are a dozen enhancements to the SURVEY command, some minor and some quite major. The preliminary documentation for these features is now available in Acrobat format. New subcommands include SKIP.RULES, OMIT.PERCENTS, ADD, SUBTRACT, MULTIPLY and DIVIDE. In addition there is more flexibility in selecting columns for significance testing.

New Features in P-STAT Version 2.21

P-STAT version 2.21 contains all the incremental changes and fixes that have been made since version 2.20 was released in 1998. The release of the manual in Acrobat format with a master index and the enhancements to the SURVEY command are major improvements. Like version 2.20, version 2.21 is Y2K compliant.

With release 2.21 the installation on Windows and NT has been simplified and a new procedure PSTATINI has been provided which allows the various environment variables to be defined and changed as needed after installation. Most of the variables that were previously set as environment variables in the AUTOEXEC.BAT file are now stored in a PSTAT.INI file in the windows directory. This means that you do not have to reboot after installation to have you PSKEY take effect.

The dBase file interface has been improved allowing larger records, more variables and retaining data types.

The external variables PSFILES, PSAUTO, PSDATA and PSTEMP make it easier for users to locate their data files.

Additional new sizes are available. Whopper 7 allows 250,000 variables per file and has a 60,000,000 byte workspace.

MAKE.CHARACTER permits numeric variables to be converted into character variables of designated sizes. It also allows character variables to be grown or shrunk. Use HELP MAKE.CHARACTER $ for details on usage.

MAKE.NUMERIC permits character variables to be converted into numeric variables without the need for PPL and renaming variables. Use HELP MAKE.NUMERIC $ for details on usage.

The sort algorithms in P-STAT have been changed to use the Quick Sort technique. File management commands such as LOOKUP and JOIN, which check for duplicate variable names, will be faster when the input files have a great many variables.

New features: April 21, 1998: Use HELP NEWS $ to check on the new commands which make it easier to locate your files if you run from a Windows program manager. PSFILES and PSDATA join PSAUTO and PSTEMP as commands which can also be installed as environment variables.; The commands which interface with dBase format files have been replaced and enhanced. The new files are named DBASE.IN and DBASE.OUT. Check the help file for details.
New features: December, 1997: List can now have value labels for character variables. The size of value labels has been increased to 80 characters and labels can now be folded.; Major enhancements to the TEXTWRITER command include value labels for character variables and full PostScript support

When PostScript is used with TEXTWRITER, the composition of a line is determined by the font that is being used and the settings of the margins in inches rather than the output width.

Underlining, fonts, and colors can be used within sentences and para paragraphs even with justify turned on. However, the justification is optimized when there are uninterupted long lines. The justification is done by apportioning the extra space equally to the blanks between words. When the line is interrupted by a font change, underline, or color change, this algorithm is also interrupted.

Tabular data can be placed left justified, right justified or centered at specific inch locations. The centering can be based on the position of a specified character such as and equal sign. These fields can be optionally underlined.

X1/Y1 and X2/Y2 coordinates can be specified to determine the location of lines and boxes.; The maximum input record size for BUILD and the output record size for FILEOUT have been greatly increased.; The DATA.ENTRY command now has a full screen frontend for free and fixed format BUILD. The frontend creates a definition file that can be saved. The BUILD command is automatically invoked when the definitions are complete.
Macro Enhancements: Macros can have arguments, either keyword or positional. Default values can be supplied for these arguments. Macros can be inserted within commands or subcommands, or can be entire blocks of commands.; There are two types of macros, BLOCK macros and INSTREAM macros. A BLOCK macro contains one or more commands which may have subcommands and data records. It is invoked by using the RUN command.; An INSTREAM macro can contain whatever one wishes. Its contents are inserted into a command or subcommand wherever !!macname is found (where macname is the name of the macro).; Both types of macro can have positional or keyword arguments, and can be defined with default values for those arguments. Alternatively, a macro can be defined without any arguments. The commands within a BLOCK macro can contain INSTREAM macro calls. INSTREAM macros can contain other instream macro calls.