Tensions in Practice: Convenient numeric rows in tension with informative gaps¶
An invented counter with a stated range rule
An invented counter reports the integers 0 through 9 and leaves a blank whenever the true count is greater than 9. Assume that is the only cause of a blank here. A numeric-only extract is easy to process, but dropping blank rows removes evidence of above-range counts. Keeping the blank as “greater than 9” retains its meaning without pretending to know the missing number.
Keep a simple numeric extract
Process recorded values with tools that expect numbers.
Retain the gap’s condition
Keep above-range observations distinct from reported zero and other values.
Why these aims pull against each other
Dropping the gap simplifies the record but changes which observation states remain available for inference.
Choose an arrangement to see what changes and what remains difficult.
Arrows express the declared relations, not measured effect sizes. Examples and quantities are illustrative.
What this choice protects
What it costs
When it fits
Compare the arrangements
Extract only numeric rows
Keep recorded numbers and omit blank rows from this extract.
- What it protects
- Numeric-only tools can process the remaining values directly.
- What it costs
- The extract loses above-range occurrences and cannot describe the entire observation process.
- When it fits
- The question explicitly concerns only reported numeric values, and omissions remain documented elsewhere if needed.
Illustration note: This is not a defensible way to estimate all counts while pretending no rows were excluded.
Keep the above-range state
Represent each blank as a bound: the count exceeded 9.
- What it protects
- The analysis retains which observations were above the instrument’s range.
- What it costs
- Downstream methods must handle bounds instead of treating every row as an exact number.
- When it fits
- The reporting mechanism is known and blanks are not also caused by transmission or other failures.
Illustration note: The source supplies the missingness-process principle; this censoring rule is a deliberately exact toy.
What this illustration does—and does not—establish
The source supplies the stated tension; the selected arrangements are bounded editorial illustrations. Costs and conditions remain part of the comparison.
- If a blank can come from another cause, the above-9 conclusion needs revision.
- Retaining the condition does not reconstruct the exact count.
- Substituting zero would falsely equate above-range counts with true zero under this model; no imputation method is supplied.
Source entries
Absence as Information
This source passage supplies the contextual tension. The concrete arrangements and schematic examples are editorial illustrations, not measured findings.
Modelled Absence versus Defaulted/Imputed Absence (measurement)
T2 — Modelled Absence versus Defaulted/Imputed Absence (measurement). The model-the-absence invariant demands gaps be treated as observations, not silently filled. The competing default — impute a value, drop the row — is standard data hygiene that here destroys signal. The failure mode is pipeline-level imputation that replaces an informative gap with a mean or zero before the analyst ever sees it. Diagnostic: ask what the data pipeline does with missing values upstream of analysis, and whether the absence-process was modelled or the gap was patched away as noise.
The source operation
Absence as information is the structural pattern in which the non-occurrence, non-presence, or non-report of something is itself diagnostic — not noise to be ignored, not a default to be filled in, but a positive signal about an underlying process. Each instance has the same shape: an *expectation* (this thing should have occurred, or normally does), an *observed gap* (it did not), and an *inference* that the gap is informative — usually a constraint on the underlying process derived from the gap, sometimes the identification of a hidden mechanism that *caused* the absence.