LoginSign Up
← Back to all sparks
P

prova

INFRA · APIS
Velocity6.3

Nonparametric Probabilistic-Statistical Variate Analysis

github.com/pglpm/prova ↗

prova adds expected-utility calculation on top of its Bayesian inference core.

r-packagesbayesian-inferencedecision-analysisapi-consolidationstatistics
◆Current state
prova does Bayesian nonparametric inference in R — probabilities through Pr() and qPr(), mutual information, quantile plots. Five releases in about two weeks renamed its central argument, collapsed two plotting functions into one, and then in v2.3.0 introduced exputility() for expected utilities and their revisability, with plot() and print() methods attached from the start.
◆Where it's heading
Two arcs run in parallel. One compresses the API: learnt= became K=, flexiplot() and plotquantiles() merged into pplot(), and omitting arguments such as Y=, X= and K= got simpler. The other extends reach — mutualinfoF() for finite-domain variates, quantile accuracy reported alongside mutual information, and now a decision-theoretic layer sitting on the inference the package already did.
◆Prediction
exputility() shipping with print() and plot() methods matches how the probability and mutual-information classes were treated, so utilities are likely to get the same class-based handling as they mature. The notes do not say whether decision analysis extends beyond expected utility.

◆Recent moves

  1. 13d ago

    exputility() brings decision analysis into prova

    ⚡ SPARK

    The release that widens prova from inference to decision-making: exputility() computes expected utilities and their revisability, arriving with plot() and print() methods rather than as a bare function. mutualinfoF() also lands as a faster path for finite-domain variates.

    View source ↗
  2. 20d ago

    Cumulative 2.x notes, plus hist() and mutual-information changes

    Explicitly a cumulative restatement of the v2.0.0 notes since v1.9.0, with two genuinely new items on top: hist() methods got a new display style and mutualinfo() now also reports quantile accuracy. The API renames it lists were already delivered in v2.0.0.

    View source ↗
  3. 23d ago

    learnt= becomes K=; pplot() replaces two plot functions

    The consolidation release: the common learnt= argument shortens to K=, and flexiplot() and plotquantiles() merge into pplot() with type = 'qx' reproducing the old behaviour. Old objects still work with the new argument, so the major version marks a tidier surface rather than a break.

    View source ↗
  4. 27d ago

    Fix for pre-existing parallel clusters

    A single bug fix for use of an already-running parallel cluster, issued four hours after v1.9.0. It is folded into every cumulative note that follows.

    View source ↗
  5. 28d ago

    extraDistr dropped; mutual-information objects get a class

    Sheds the extraDistr dependency, gives mutual-information results their own "mi" class with a print() method, and reworks digit rounding and probability display. The class work here is what the v2.3.0 utility objects later build on.

    View source ↗