Estimate uncertain quantities or examine how analytical choices affect results.
Method toolkit R
Estimate a quantity by breaking it into smaller uncertain components and simulating possible outcomes.
Define the component distributions and how they combine. Results depend on your assumptions; they are probabilistic estimates rather than observed totals.
Method toolkit R
Test how theoretically plausible analytical choices affect statistical results using specification curves.
This is Humaniverse’s fork of specr; the linked documentation describes the upstream methods. Choose defensible specifications. A specification curve does not make an estimate more valid or generalisable.
App / Method toolkit R, Shiny, specr
Interactively explore a specification curve, its analytical choices and the individual models behind it.
Requires a specr result object and local R/Shiny dependencies. Supplying the original dataset enables model refitting. No hosted app is documented.
Choose the method that matches the uncertainty you want to examine. Use fermi for uncertain input quantities, or specr and shinyspecr for analytical choices in statistical models. A specr result can be passed to shinyspecr; other combinations are not an implied pipeline. Record the assumptions tested and the uncertainties that remain outside the analysis.
App / Method toolkit R, Shiny
For a composite index built with COINr, explore sensitivity to weights, normalisation and aggregation within shinycoin.
Requires a local R/Shiny setup. This is index-specific sensitivity analysis, separate from the statistical specification curves supported by specr.
