dc
0ee0b4ea7c
The validator fix lands in dc/dc (PR #323), so the post describes the fixed behaviour rather than the mismatch: a special missing fails NOTNULL, matching a physical SAS NOT NULL / primary key constraint. Also records what ROUND and SOFTSELECT/HARDSELECT do with one, and groups MINVAL/MAXVAL/HARDFORMULA/SOFTFORMULA under 'not supported'.
142 lines
11 KiB
Markdown
142 lines
11 KiB
Markdown
---
|
|
title: '28 Ways to Be Missing in SAS'
|
|
description: A SAS numeric missing is not a lone wolf - there are 28 of them, and Data Controller has handled all of them since v4. How they work, and what the validation rules do with them.
|
|
date: '2026-09-22 09:00:00'
|
|
author: 'Allan Bowe'
|
|
authorLink: https://www.linkedin.com/in/allanbowe/
|
|
tags:
|
|
- Special Missings
|
|
- Data Quality
|
|
previewImg: './cover.jpeg'
|
|
---
|
|
|
|
# 28 Ways to Be Missing in SAS
|
|
|
|
A numeric missing value in SAS is not a lone wolf. The ordinary missing (`.`) is one of **28** distinct missing values available for a numeric variable - the other 27 are written with a single character, the letters `A` to `Z` or an underscore (`._`).
|
|
|
|
They exist because "missing" is usually not the whole story. A survey question that was never reached, a reading that was illegible, a value the respondent refused to give - in a well run process those are different facts, and a lone `.` throws the difference away. Special missings record *why* the value is missing.
|
|
|
|
## They are numbers, not text
|
|
|
|
- Arithmetic on them yields missing - `.A + 1` is `.`
|
|
- `proc means`, `proc summary` and friends exclude them, exactly as they exclude `.`
|
|
- They sort below every non-missing number, in the order `._`, then `.`, then `.A` to `.Z`
|
|
- The `NMISS()` and `CMISS()` functions count them as missing
|
|
- Converting one to text drops the period - `cats(.A)` is the string `A`
|
|
|
|
A display gotcha worth knowing: a regular missing prints as `.`, unless `options MISSING=` changes that character - set it to blank and a regular missing renders as an empty cell. The option affects only the regular missing; `._` and `.A`-`.Z` always print as their own letter. So a blank cell in a SAS listing is still unambiguously a regular missing, and a lone letter is still a special one. Data Controller itself is unaffected either way - it sends a regular missing to the browser as `null` and a special missing as its letter.
|
|
|
|
In a SAS dataset they are written with a leading period (`.A`, `.B` ... `._`). In Data Controller you type just the letter or the underscore - no period - and the letter is not case sensitive. Two letters, or a letter mixed with a number, are refused rather than guessed.
|
|
|
|
## Carrying them between the browser and SAS
|
|
|
|
The Data Controller frontend and the SAS backend exchange data as JSON, and JSON has no way to express a letter as a numeric value - `A` is a string. The conversion is handled in the open source [SASjs Adapter](https://github.com/sasjs/adapter#variable-types):
|
|
|
|
- The adapter infers each column's SAS type from the values it is given. All numeric values means numeric; all strings means character; and a column holding a single character (`a`-`z`, `_` or `.`) alongside numeric values is numeric, with the lone characters written to SAS as special missings.
|
|
- `null` becomes `.` or an empty string, according to the type derived for that column.
|
|
- Two cases cannot be inferred from the values alone: a numeric column containing *only* special missings looks like a single character column, and a character column containing only nulls looks numeric. For those, the adapter accepts an explicit format for the column - and Data Controller sends one automatically, because it already knows each column's SAS format from the metadata returned by the backend.
|
|
- A value that is neither a number nor a single valid character is refused rather than guessed (for example `aaaa`, or `!` in a numeric column), and a literal `.` is refused in favour of `null` for a regular missing.
|
|
|
|
There is nothing to configure. Special missings are available by default, for numeric cells.
|
|
|
|
Once in SAS they are ordinary values, so they are what the approval DIFF screen compares, and the DIFF's formatted / unformatted switch shows either the formatted representation or the raw value - useful for confirming exactly which missing was set.
|
|
|
|
## What the Data Controller validation rules do with them
|
|
|
|
These are Data Controller's own rules, configured per column in the `MPE_VALIDATIONS` table and applied in the browser as you edit and submit.
|
|
|
|
- `NOTNULL` - fails. A special missing is NULL as far as a NOT NULL (or primary key) constraint is concerned, so the rule rejects it - and so does a physical SAS NOT NULL constraint on the target table
|
|
- `HARDREGEX` - checked against the pattern like any other value; unlike blanks and the plain `.`, special missings are **not** exempt, so a numeric column that carries them needs a pattern which allows for a single letter
|
|
- `SOFTREGEX` - the same check, but a failure is only a warning rather than a block, and it is ignored entirely if the column also has a `HARDREGEX` rule
|
|
- `SOFTSELECT` - never blocks, so a special missing passes; `HARDSELECT` is a strict membership test and rejects one
|
|
- `ROUND` - harmless. It only rounds values that are numbers, so a special missing is left exactly as it was typed
|
|
|
|
Special missings are **not supported** in a column that carries a range rule or a formula:
|
|
|
|
- `MINVAL` / `MAXVAL` - a special missing sorts below every number, so `MINVAL` rejects it and `MAXVAL` accepts it while rejecting the column's real numbers. The rule's own value has to be a number too: put a special missing in `MINVAL` and every cell in the column fails, put one in `MAXVAL` and every real number fails
|
|
- `HARDFORMULA` / `SOFTFORMULA` - a formula that reads a special-missing cell does not compute; the grid's spreadsheet engine returns `#VALUE!` for that row, where the plain `.` contributes 0
|
|
|
|
`CASE` (`UPCASE` / `LOWCASE`) is a character rule and does not come into it: SAS hands the browser a special missing as an uppercase letter, so there is no case to enforce, and a `CASE` rule on a numeric column would reject the column's numbers rather than the missing.
|
|
|
|
One thing worth knowing about the regex rules: although the pattern is written in SAS PRX syntax, the check itself runs in the browser - SAS only parses the pattern (`PRXPARSE`) when the rule is saved - so the two engines can disagree on exotic patterns, and SAS pads a numeric-to-character conversion, so a pattern re-used in SAS needs `strip()` for an anchored match.
|
|
|
|
In short, a special missing counts as a value for the pattern and dropdown rules, and as a missing value for everything that asks for a number.
|
|
|
|
`video: [Managing Special Missing Values with Data Controller for SAS](https://www.youtube-nocookie.com/embed/ggrcNr23Jzw)`
|
|
|
|
The recording shows the full cycle: typing special missings into numeric cells, the rejections, submitting the changes, approving them, and reviewing the DIFF - including a change from one special missing to another, and the formatted / unformatted switch on date and datetime columns.
|
|
|
|
<!--
|
|
LinkedIn version of this post - publish it with the "Managing Special Missing
|
|
Values with Data Controller for SAS" video attached. Kept in sync with the copy
|
|
above: same points, same claims, same order.
|
|
|
|
A SAS numeric missing is not a lone wolf. It is 28.
|
|
|
|
The ordinary missing (.) is just the most common of them. The other 27 are single characters - the letters A to Z, or an underscore (._) - and they let you record WHY a value is missing: the question was never reached, the reading was illegible, the respondent refused.
|
|
|
|
They are real numeric values, not text:
|
|
|
|
- .A + 1 is .
|
|
- PROC MEANS excludes them, like any other missing
|
|
- they sort below every number: ._ then . then .A to .Z
|
|
- NMISS() and CMISS() count them as missing
|
|
- a regular missing prints as . unless options MISSING changes it - and that option never touches a special missing
|
|
|
|
The hard part is every tool that is not SAS. JSON has no way to say "this number is a letter" - A is just a string. So a value that is perfectly legal in a SAS dataset quietly breaks in the browser, in Excel, in an API.
|
|
|
|
We solved that in the open source SASjs Adapter, and it has been in Data Controller for SAS since v4:
|
|
|
|
- the adapter infers the column type from the values, so a column of numbers that also holds a lone letter is written to SAS as numeric, with the letter as a special missing
|
|
- where the type cannot be inferred - a numeric column holding ONLY special missings - Data Controller passes the column format explicitly
|
|
- you type the letter on its own: a or A, no period, and case does not matter
|
|
|
|
How they behave in Data Controller's validation rules:
|
|
|
|
- NOTNULL fails - a special missing is null to a NOT NULL (or primary key) constraint, and a real SAS NOT NULL constraint rejects it too
|
|
- HARDREGEX checks it against the pattern, unlike blanks and plain .
|
|
- SOFTREGEX warns instead of blocking (and is ignored if the column also has a HARDREGEX)
|
|
- SOFTSELECT lets it through; HARDSELECT rejects it
|
|
- ROUND leaves it alone, because it only rounds numbers
|
|
|
|
Special missings are not supported where a range rule or a formula is configured:
|
|
|
|
- MINVAL rejects it; MAXVAL accepts it but rejects the column's real numbers
|
|
- a HARDFORMULA/SOFTFORMULA that reads a special-missing cell returns #VALUE! rather than a number
|
|
|
|
In short: a value for the pattern and dropdown rules, a missing value for anything that wants a number.
|
|
|
|
The video shows the whole cycle - entering them, the rejections, submit, approve, and the DIFF.
|
|
|
|
#sas #datacapture #mdm #dataquality
|
|
|
|
Image prompt: Cinematic editorial cover illustration: a large pack of wolves, a
|
|
dozen or more, spread wide and moving together across a vast snow plain at dusk,
|
|
seen from a low wide angle. One wolf stands apart from the group on the left of
|
|
frame, turned back toward the pack. The pack is rendered in cool blue-grey and
|
|
silver; the lone wolf catches the only warm light in the scene, a single low
|
|
amber sun. Overcast dusk sky, faint falling snow, long soft shadows, generous
|
|
empty sky and snow to the upper third so the wide crop breathes. Painterly
|
|
digital illustration, muted desaturated palette, soft depth of field with the
|
|
distant wolves falling out of focus, no text, no letters, no numbers, no logos,
|
|
no watermark.
|
|
|
|
Constraints: no text, no letters, no numbers, no digits, no symbols, no
|
|
captions, no logos, no watermark, no signature, no border, no frame, no collar,
|
|
no harness, no humans, no buildings, and do not ask for an exact head count of
|
|
28 - a crowded pack reads worse than a dozen clear animals, and the number
|
|
belongs in the headline, not the artwork. The no-text rule matters more than
|
|
usual here: the subject is letters standing in for numbers, so a stray glyph
|
|
anywhere undercuts the cover.
|
|
|
|
Output: save as ./cover.jpeg at 1.91:1 (1200x627, matching the other feed
|
|
covers and doubling as the LinkedIn share card), and set
|
|
previewImg: './cover.jpeg' in the front matter above. Do not also embed the
|
|
image in the markdown body. Full prompt, variants and rationale:
|
|
https://paste.4gl.io/?513711f5bde3610e#9LZCA9xzQHDy3CycXmsWU5AoVNtjBZmPotUHHtHTAerF
|
|
|
|
Fallback variant if the pack composition comes back muddled: a single wolf
|
|
standing alone on a snow plain at dusk, its shadow stretching toward a distant
|
|
pack reduced to small silhouettes on the horizon.
|
|
-->
|