design/user-stories/api_requirements.md

In FPRgroup/FPEM: FPEM

API requirements

Data management

• Maintain "official" versions of core data (e.g. survey, demographics, ISO country)
• Allow "official" versions of data to be retrieved/queried
• Allow input data to be updated
• Allow model outputs to circle-back for runs on subsets (this may not require anything new).

Core modeling

• Allow user-defined models matching pre-defined input/output scenarios
• Run models, varying the following parameters
  • Type---i.e. married, unmarried, all women
  • Age range
  • Time frame
  • Prevalence dataNumber of women
  • Prior parameters

• Allow user-defined lists of arguments to be passed down to the algorithms

• Provide high-level parameter aggregations that make it easier to run "standard" analyses.

• Obtain uncertainty around estimates (e.g. in the form of posterior marginal quantiles)

• Allow results and computationally expensive intermediate data to be extracted/stored and re-loaded. Provide access to the input data used to run the model.

Model and algorithm checking

• cross-validation
• out-of-sample validation

Model validation

• Run out-of-sample validations for a data/model/algorithm combo.
• Create goodness-of-fit metrics of times-series for multiple units and metrics
• Compare runs for different parameters

Aggregation

• Create country (or unit)-level aggregates
• Create time-aggregates for given time-periods
• Create standard aggregates (following, e.g. UNPD procedure, or a particular paper).

Reporting

• Create standard summaries of estimated time-series
• Create standard table (e.g.-as published by X)

Visualization

• Compare results for arbitrary pairs and triples of years probabilistically.
• Review uncertainty around estimates (e.g. in the form of posterior marginal quantiles)
• Generate web-friendly and print-friendly and screen-friendly viz. from the same basic plots/data.

FPRgroup/FPEM documentation built on March 3, 2020, 8:19 a.m.