src/procedures/procedures.lisp

   1 ;;; -*- mode: lisp -*-
   2
   3 ;;; Time-stamp: <2009-10-09 12:17:47 tony>
   4 ;;; Creation:   <2009-03-12 17:14:56 tony>
   5 ;;; File:       procedures.lisp
   6 ;;; Author:     AJ Rossini <blindglobe@gmail.com>
   7 ;;; Copyright:  (c)2009--, AJ Rossini.  Currently licensed under MIT
   8 ;;;             license.  See file LICENSE.mit in top-level directory
   9 ;;;             for information.
  10 ;;; Purpose:    Classes for statistical procedures, and generics
  11 ;;;             supporting use of such procedures.
  12
  13 ;;; What is this talk of 'release'? Klingons do not make software
  14 ;;; 'releases'.  Our software 'escapes', leaving a bloody trail of
  15 ;;; designers and quality assurance people in its wake.
  16
  17 ;;; This organization and structure is new to the 21st Century
  18 ;;; version..   Think, "21st Century Schizoid Man".
  19
  20 (in-package :cls-statproc)
  21
  22 ;;; Statistical procedures can consist of one or more of:
  23
  24 ;;; - mathematical forms: linear predictors, splines, functional
  25 ;;;   forms) which get spliced together.
  26
  27 ;;; - criteria functions (likelihoods, sum-of-squares, information), params and data.
  28 ;;; - optimization algorithmsfunctions
  29
  30 ;;; - root functions (score functions, influence functions), params and data
  31 ;;; - zero-finding algorithms
  32
  33 ;;; - point estimate: special case of...
  34 ;;; - interval estimation (with uncertainty criteria to support range or point)
  35 ;;; - hypothesis test -- which can be thought of as a point estimator
  36 ;;;   for a coarsened problem, or as a simple interval-based decision with
  37 ;;;   uncertanty criteria.
  38
  39 ;;; Estimators are reaasonably straightforward.  Either we are
  40 ;;; providing a point estimate, or we provide an interval estimate.
  41 ;;; But shouldn't these be characterized in the same manner?
  42
  43 ;;; We are currently building up an ontology (ADG/DAG) or
  44 ;;; knowledge-base whose leaves are described by the path to them.
  45
  46 (defvar *statistical-procedure-components*
  47   '((interval-estimation
  48
  49      (hypothesis-testing ; coarsened interval
  50       (frequentist
  51        (fisherian
  52         (neyman-pearson
  53          (wald-test
  54           t-test
  55           likelihood-ratio-test
  56           score-test))))))
  57     (point-estimation
  58      maximum-likelihood-estimation)
  59
  60     ()
  61     optimization
  62     root-finding
  63
  64     criteria-functions
  65
  66     mathematical-forms
  67     ))
  68
  69 (defclass statistical-ontology-class ()
  70   ()
  71   (:documentation "container class for manipulation of structural components."))
  72
  73 (defclass statistical-decision ()
  74   ((ontology-classification :initform nil
  75                             :initarg :ontology-classification
  76                             :type statistical-ontology-class
  77                             ))
  78   (:documentation "instance describing the end result, if it is an
  79   interval/range/region, or point estimate, or a conclusion from a
  80   test (i.e. hypothesis(es) selected, strength of conclusion)"))
  81
  82 (defclass statistical-dataset (dataframe-like)
  83   (:documentation "a particular dataset, usually the subset from a
  84   larger set, which is used as the input to the procedure."))
  85
  86 (defclass statistical-metadata (dataframe-like)
  87   (:documentation "the description of the dataset's statistical
  88   properties which are required for the procedure to work or meet
  89   assumptions."))
  90
  91 (defclass statistical-procedure ()
  92   ((ontological-spec :initform nil
  93                      :initarg :ontology-def
  94                      :type list
  95                      :accessor ontology
  96                      :documentation "list of symbols describing
  97   ontological classification")
  98    (how-to-fit)
  99    (how-to-simulate)
 100    (instance-data)))
 101
 102 ;;;
 103 (defgeneric proc-consistents-data-p (metadata data)
 104   (:documentation "verify that the metadata required for a procedure
 105   is present in a particular dataset.  The dataset will usually be a
 106   subset of the full working dataset."))
 107
 108 (defgeneric process-data (proc data)
 109   (:documentation "Run the statistical procedure on the dataset and
 110   report the decision."))