Time-series data | Data-frame model | Distribution fitting
InputData is the second scientific collection in the project tree. It owns the DataFrame supplied to fitting and univariate analyses and provides the workflow that converts manual, agency, block-series, or peaks-over-threshold input into exact, interval, threshold, or uncertain observations.
Exact data may be entered manually, derived as a block series, derived as a peaks-over-threshold series, or downloaded as USGS peak discharge or peak stage. When input is derived from TimeSeriesElement, the configuration records the block function and window, start/end months, smoothing rule and period, threshold, and minimum separation between peaks.
The resulting DataFrame is the scientific handoff to the model library. Its chronology, censoring semantics, observation indexes, perception thresholds, plotting positions, and uncertain-observation distributions are defined in the data-frame chapter.
- A block series selects the configured statistic within each declared time block.
- A peaks-over-threshold series applies the configured threshold and minimum inter-peak separation to the source chronology.
- Multiple Grubbs-Beck processing is optional and applies only in the supported Bulletin 17C workflow.
IsProcesseddistinguishes a configured source from a completed derived data frame.- Changes to the source time series or processing controls clear dependent results and require reprocessing.
Opening a saved project repairs HS plotting positions in valid frames containing explicitly recorded values below their covering perception thresholds (issue #19). These recorded values are ranked by magnitude while threshold-only years retain their censored weighting. Changed positions mark the element dirty and are persisted by normal saving; source observations, perception thresholds, and plot settings are preserved. Invalid frames remain loadable for correction through the usual validation workflow.
The repair updates the input frame. Saved analyses are not automatically re-estimated. Reprocess affected analyses to refresh empirical moments, regression-on-order-statistics initialization, and fit diagnostics that depend on the corrected positions. See plotting positions and derived state for the calculation and direct XML serialization contracts.
The element validates its source reference, processing window, threshold/separation controls, data-frame contents, and series-specific requirements. Processing does not establish stationarity, independence, representativeness, perception-threshold correctness, or fitness for a particular frequency model; those remain model and application assumptions.
Implementation is in RMC.BestFit.UI/Elements/InputData/InputData.cs and InputDataCollection.cs. Fast UI tests cover construction, dirty-state propagation, collection persistence, validation, processing state, and project lifecycle. The numerical report treats data preparation as a controlled input boundary and verifies scientific calculations in the analysis that consumes the resulting frame; see Input-Data Verification Boundary.