Differences between revisions 8 and 15 (spanning 7 versions)

SPSS Data Types

SPSS exposes numeric and string data types.

Certain specialized forms of data are handled by formats that impact visualization and export, not storage. The primary example of this is date and time data.

Contents

SPSS Data Types

Numeric Data

Numeric data is stored as double-precision floating point.

Data formats adjust the visualized/exported representation of a data point, often only truncating the value. Adjusting these data formats does not destroy precision.

Given the literal value 123.45...

General Format	Format	Representation
Fw	F8	123
Fw.d	F8.1	123.4
Nw	N8	00000123
Nw.d	N8.1	000123.4

See here for the built-in library of numeric functions.

Date and Time Formats

Date data is stored as the number of seconds from midnight, October 14, 1582 to midnight on the specified date.

Datetime data is stored as the number of seconds to the specified time on the specified date.

Time data is stored as the number of seconds. This can be imagined as the number of seconds from midnight to the specified time, but that is not a necessary construct.

Keep in mind that 1 day = 60 (seconds) * 60 (minutes) * 24 (hours) = 86400 seconds.

Format	Representation
`DATETIME20`	`dd-MMM-yyyy hh:mm:ss`
`DATETIME17`	`dd-MMM-yyyy hh:mm`
`DATE11`	`dd-MMM-yyyy`
`DATE9`	`dd-MMM-yy`
`TIME8`	`hh:mm:ss`
`TIME5`	`hh:mm`
`ADATE10`	`mm/dd/yyyy`
`EDATE10`	`dd.mm.yyyy`
`SDATE10`	`yyyy/mm/dd`

Note that MMM appears as JAN, not 001.

Note that DATE11 and DATE9 differ in the representation of years (4 and 2 digits respectively). The American, European, and Sortable date formats have a similar feature by swapping [AES]DATE10 with [AES]DATE8.

See here for the built-in library of date and time functions.

String Data

String data is stored at a fixed length. This length is defined and adjusted through the format; Aw where w is the width.

String Formats

To explicitly re-size a string, try:

alter type VAR (a100).

To re-size a string to the smallest possible length (without losing data), try:

alter type VAR (a=amin).

The size does not automatically grow to accept longer values, and string expressions do not automatically strip trailing whitespace. Consider the below example:

data list list / numeric_zip_code (F5).
begin data
12345
1234
123
end data.

string padding (A2).
if numeric_zip_code lt 10000 padding="0".
if numeric_zip_code lt 1000  padding="00".
string zip_code (A5).
compute zip_code=concat(padding, string(numeric_zip_code,F5)).
execute.

This will produce:

numeric_zip_code	padding	zip_code
`12345`	`" "`	`" 123"`
`1234`	`"0 "`	`"0 123"`
`123`	`"00"`	`"00123"`

The concat function is returning a 7-long string value, and SPSS is silently truncating it to 5-long before storing in zip_code.

There are better ways to zero-pad a string, but fundamentally the mistake is not wrapping string expressions in ltrim and rtrim calls.

See here for the built-in library of string functions.

String Literals

String literals are declared by wrapping a string value in quotes, either single (') or double (").

Quote marks within strings can be handled in one of two ways: either use the opposite quote mark to define the string, or escape the quote mark.

variable label var1 "Say ""Hello!""".
variable label var2 'Don''t say "Goodbye!"'.

CategoryRicottone

-  ⇤ ← Revision 8 as of 2020-07-04 14:40:13 → 
  Size: 3523
  Editor: DominicRicottone
  Comment:
+   ← Revision 15 as of 2023-01-13 20:22:58 → ⇥
  Size: 4215
  Editor: DominicRicottone
  Comment:
-Deletions are marked like this.
+Additions are marked like this.
 Line 1:
-## page was renamed from SPSSFormats
= SPSS Formats =
SPSS supports numeric variables and strings, with conversion options and some automatic conversion functionality. There are several formats on top of numerics that impact visualization, not storage.
+= SPSS Data Types =

SPSS exposes numeric and string data types.

Certain specialized forms of data are handled by formats that impact visualization and export, not storage. The primary example of this is date and time data.
-Line 10:
+Line 12:
+== Numeric Data ==
-Line 11:
+Line 14:
-== Datetime ==
+'''Numeric''' data is stored as double-precision floating point.
-Line 13:
+Line 16:
-=== A warning about data input ===
+Data formats adjust the visualized/exported representation of a data point, often only truncating the value. Adjusting these data formats does not destroy precision.
-Line 15:
+Line 18:
-SPSS tries to be clever about reading data.
+Given the literal value 123.45...
-Line 17:
+Line 20:
-Under the format `TIME`, all of these are read in as `"01:02"`, even though the format is a minimum of 5-wide:
 * `"1:2"`
 * `"01 2"`
 * `"01:02"`
+||'''General Format''' ||'''Format'''   ||'''Representation''' ||
||Fw                   ||F8             ||123                  ||
||Fw.d                 ||F8.1           ||123.4                ||
||Nw                   ||N8             ||00000123             ||
||Nw.d                 ||N8.1           ||000123.4             ||
-Line 22:
+Line 26:
-Under the format `DATE`, all of these are read in as `"28-OCT-90"`:
 * `"28-OCT-90"`
 * `"28/10/1990"`
 * `"28.OCT.90"`
 * `"28 October, 1990"`
+See [[SPSS/NumericFunctions|here]] for the built-in library of numeric functions.
 Line 30:
-=== Datetime formats ===
+=== Date and Time Formats ===
 Line 32:
-||'''Format''' ||'''Appears as...'''    ||'''Note'''              ||
||DATETIME20   ||`dd-MMM-yyyy hh:mm:ss` ||`MMM` is as JAN, not 001||
||DATETIME18   ||`dd-MMM-yyyy hh:mm`    ||`MMM` is as JAN, not 001||
||DATE11       ||`dd-MMM-yyyy`          ||`MMM` is as JAN, not 001||
||DATE7        ||`dd-MMM-yy`            ||`MMM` is as JAN, not 001||
||TIME8        ||`hh:mm:ss`             ||                        ||
||TIME5        ||`hh:mm`                ||                        ||
||ADATE10      ||`mm/dd/yyyy`           ||                        ||
||ADATE8       ||`mm/dd/yy`             ||                        ||
||EDATE10      ||`dd.mm.yyyy`           ||                        ||
||EDATE8       ||`dd.mm.yy`             ||                        ||
+'''Date''' data is stored as the number of seconds from midnight, October 14, 1582 to midnight on the specified date.
-Line 44:
+Line 34:
+'''Datetime''' data is stored as the number of seconds to the specified time on the specified date.
-Line 45:
+Line 36:
+'''Time''' data is stored as the number of seconds. This ''can'' be imagined as the number of seconds from midnight to the specified time, but that is not a ''necessary'' construct.
-Line 46:
+Line 38:
+Keep in mind that 1 day = 60 (seconds) * 60 (minutes) * 24 (hours) = 86400 seconds.
-Line 47:
+Line 40:
-=== Parse Timestamps ===
+||'''Format''' ||'''Representation'''   ||
||`DATETIME20` ||`dd-MMM-yyyy hh:mm:ss` ||
||`DATETIME17` ||`dd-MMM-yyyy hh:mm`    ||
||`DATE11`     ||`dd-MMM-yyyy`          ||
||`DATE9`      ||`dd-MMM-yy`            ||
||`TIME8`      ||`hh:mm:ss`             ||
||`TIME5`      ||`hh:mm`                ||
||`ADATE10`    ||`mm/dd/yyyy`           ||
||`EDATE10`    ||`dd.mm.yyyy`           ||
||`SDATE10`    ||`yyyy/mm/dd`           ||
-Line 49:
+Line 51:
-A macro for timestamp conversion, from `dd/mm/yyyy hh:mm` to SPSS `DATETIME20`:
+Note that `MMM` appears as `JAN`, not `001`.
-Line 51:
+Line 53:
-{{{
define !str2datetime (in=!tokens(1) / out=!cmdend)
  numeric !out (DATETIME20).
  compute !out =
   date.mdy(
    number(char.substr(!in,1,2), f2),
    number(char.substr(!in,4,2), f2),
    number(char.substr(!in,7,4), f4)
   )
   + time.hms(
    number(char.substr(!in,12,2), f2),
    number(char.substr(!in,15,2), f2),
    0
   )
  .
!enddefine.
+Note that `DATE11` and `DATE9` differ in the representation of years ''(4 and 2 digits respectively)''. The American, European, and Sortable date formats have a similar feature by swapping `[AES]DATE10` with `[AES]DATE8`.
-Line 68:
+Line 55:
-!str2datetime in=mytimestampvar out=mydatetimevar.
}}}



=== Create Timestamps ===

This is unfortunately long-winded.

{{{
data list free / a.
begin data.
1
end data.
dataset name TIMESTAMP.
execute.

string tmp_date (A11) datetime (A20).
compute tmp_date = $date.
alter type tmp_date (date11).
compute tmp_time = $time.

* Built-in formats.
compute datetime=string(tmp_date, ADATE10).
execute.

* Manual formats.
compute datetime=concat(
  char.substr(string(xdate.year(tmp_date),N4),3,2),
  string(xdate.month(tmp_date),N2),
  string(xdate.mday(tmp_date),N2),
  "_",
  string(xdate.hour(tmp_time),N2),
  string(xdate.minute(tmp_time),N2),
).
execute.

do if $casenum=1.
  write outfile 'timestamp.sps' /1 "define !timestamp() " datetime " !enddefine.".
end if.
dataset close TIMESTAMP.
execute.

insert file='timestamp.sps'.
execute.
}}}
+See [[SPSS/DateTimeFunctions|here]] for the built-in library of date and time functions.
-Line 119:
+Line 61:
-== Numerics ==
+== String Data ==
-Line 121:
+Line 63:
-Formats only truncate the displayed value, or the value translated into data files. SPSS always retains data to original precision.

Given 123.45...

||'''Format''' ||'''Using...''' ||'''Appears as...''' ||
||Fw           ||F8             ||123                 ||
||Fw.d         ||F8.1           ||123.4               ||
||Nw           ||N8             ||00000123            ||
||Nw.d         ||N8.1           ||000123.4            ||

----
+'''String''' data is stored at a fixed length. This length is defined and adjusted through the format; `Aw` where `w` is the width.
-Line 135:
+Line 67:
-== Strings ==
+=== String Formats ===
-Line 137:
+Line 69:
-String variables are only formatted to a length. The format is specified as `AN` where `N` is the width.
+To explicitly re-size a string, try:

{{{
alter type VAR (a100).
}}}

To re-size a string to the smallest possible length (without losing data), try:

{{{
alter type VAR (a=amin).
}}}

The size does ''not'' automatically grow to accept longer values, and string expressions do not automatically strip trailing whitespace. Consider the below example:

{{{
data list list / numeric_zip_code (F5).
begin data
12345
1234
123
end data.

string padding (A2).
if numeric_zip_code lt 10000 padding="0".
if numeric_zip_code lt 1000  padding="00".
string zip_code (A5).
compute zip_code=concat(padding, string(numeric_zip_code,F5)).
execute.
}}}

This will produce:

||'''numeric_zip_code'''||'''padding'''||'''zip_code'''||
||`12345`               ||`"  "`       ||`"  123"`     ||
||`1234`                ||`"0 "`       ||`"0 123"`     ||
||`123`                 ||`"00"`       ||`"00123"`     ||

The `concat` function is returning a 7-long string value, and SPSS is silently truncating it to 5-long before storing in `zip_code`.

There are better ways to zero-pad a string, but fundamentally the mistake is not wrapping string expressions in `ltrim` and `rtrim` calls.

See [[SPSS/StringFunctions|here]] for the built-in library of string functions.



== String Literals ==

String literals are declared by wrapping a string value in quotes, either single (') or double (").

Quote marks within strings can be handled in one of two ways: either use the opposite quote mark to define the string, or escape the quote mark.

{{{
variable label var1 "Say ""Hello!""".
variable label var2 'Don''t say "Goodbye!"'.
}}}

Diff for "SPSS/DataTypes"