Read the gaps as carefully as the data

Historical state#

An account may have a publication date without establishing the date of the configuration it describes. Avoid turning an undated observation into a precise chronology.

Identity and relationships#

Shared names and similar specifications can suggest a relationship without proving identity. Unresolved actor or represented-instrument links should remain unresolved in downstream analysis unless additional evidence supports a decision.

Geographic precision#

Coordinates can refer to locations at different levels of precision. Preserve coordinate qualifications and administrative evidence when making maps or spatial comparisons.

Counts and occurrences#

A source-account row is not necessarily a unique physical component. For example, multiple accounts may describe the same stop at different times. Count the unit that your question actually concerns.

Missing values have different causes#

Keep these distinctions through transformations. Converting all missing values to zero, attaching an arbitrary year or resolving every name by its nearest spelling can make a table easier to compute while changing its meaning. If a method requires complete cases, document the selection rule and assess which sources or regions it disproportionately removes.

Repeated accounts are not independent witnesses#

Agreement across records can be valuable, but sources may copy one another or derive from a shared earlier account. Count supporting records separately from independent evidence. When the transmission relationship is unknown, that uncertainty belongs in the interpretation rather than being replaced by a simple majority vote.

A bounded comparison is often more useful#

For a comparison of specifications, choose a documented unit, date scope and inclusion rule before interpreting differences. Report both what the selected corpus supports and what it cannot answer. A bounded result about attributed accounts can be strong research without claiming to describe every surviving instrument or its present condition.

One coverage statistic, carefully bounded#

For POD 1.6.0, 143,880 of 215,769 canonical organs have stop accounts: approximately 66.7 percent. This describes presence of that category of documentation in the specified release. It does not assess whether each account is complete, whether multiple accounts concern the same state, or how well the corpus represents all historical organs. The 3,604,579 public stop rows use another counting unit and must not be substituted for a count of physical stops.