• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1[section:issues Known Issues, and TODO List]
2
3Predominantly this is a TODO list, or a list of possible
4future enhancements.  Items labeled "High Priority" effect
5the proper functioning of the component, and should be fixed
6as soon as possible.  Items labeled "Medium Priority" are
7desirable enhancements, often pertaining to the performance
8of the component, but do not effect it's accuracy or functionality.
9Items labeled "Low Priority" should probably be investigated at
10some point.  Such classifications are obviously highly subjective.
11
12If you don't see a component listed here, then we don't have any known
13issues with it.
14
15[h4 tgamma]
16
17* Can the __lanczos be optimized any further?  (low priority)
18
19[h4 Incomplete Beta]
20
21* Investigate Didonato and Morris' asymptotic expansion for large a and b
22(medium priority).
23
24[h4 Inverse Gamma]
25
26* Investigate whether we can skip iteration altogether if the first approximation
27is good enough (Medium Priority).
28
29[h4 Polynomials]
30
31* The Legendre and Laguerre Polynomials have surprisingly different error
32rates on different platforms, considering they are evaluated with only
33basic arithmetic operations.  Maybe this is telling us something, or maybe not
34(Low Priority).
35
36[h4 Elliptic Integrals]
37
38* [para Carlson's algorithms (mainly R[sub J]) are somewhat prone to
39internal overflow/underflow when the arguments are very large or small.
40The homogeneity relations:]
41[para R[sub F](ka, kb, kc) = k[super -1/2] R[sub F](a, b, c)]
42[para and]
43[para R[sub J](ka, kb, kc, kr) = k[super -3/2] R[sub J](a, b, c, r)]
44[para could be used to sidestep trouble here: provided the problem domains
45can be accurately identified. (Medium Priority).]
46* There are a several other integrals: Bulirsch's ['el] functions that could
47be implemented using Carlson's integrals (Low Priority).
48* The integrals K(k) and E(k) could be implemented using rational
49approximations (both for efficiency and accuracy),
50assuming we can find them. (Medium Priority).
51
52[h4 Owen's T Function]
53
54There is a problem area at arbitrary precision when ['a] is very close to 1.  However, note that
55the value for ['T(h, 1)] is well known and easy to compute, and if we replaced the
56['a[super k]] terms in series T1, T2 or T4 by ['(a[super k] - 1)] then we would have the
57difference between ['T(h, a)] and ['T(h, 1)].  Unfortunately this doesn't improve the
58convergence of those series in that area.  It certainly looks as though a new series in terms
59of ['(1-a)[super k]] is both possible and desirable in this area, but it remains elusive at present.
60
61[h4 Statistical distributions]
62
63* Student's t Perhaps switch to normal distribution
64as a better approximation for very large degrees of freedom?
65
66[h4 Feature Requests]
67
68The following table lists distributions that are found in other packages
69but which are not yet present here, the more frequently the distribution
70is found, the higher the priority for implementing it:
71
72[table
73[[Distribution][R][Mathematica 6][NIST][Regress+][Matlab]]
74
75[/3 votes:]
76[[Geometric][X][X][-][-][X]]
77
78[/2 votes:]
79[[Multinomial][X][-][-][-][X]]
80[[Tukey Lambda][X][-][X][-][-]]
81[[Half Normal / Folded Normal][-][X][-][X][-]]
82[[Chi][-][X][-][X][-]]
83[[Gumbel][-][X][-][X][-]]
84[[Discrete Uniform][-][X][-][-][X]]
85[[Log Series][-][X][-][X][-]]
86[[Nakagami (generalised Chi)][-][-][-][X][X]]
87
88[/1 vote:]
89[[Log Logistic][-][-][-][-][X]]
90[[Tukey (Studentized range)][X][-][-][-][-]]
91[[Wilcoxon rank sum][X][-][-][-][-]]
92[[Wincoxon signed rank][X][-][-][-][-]]
93[[Non-central Beta][X][-][-][-][-]]
94[[Maxwell][-][X][-][-][-]]
95[[Beta-Binomial][-][X][-][-][-]]
96[[Beta-negative Binomial][-][X][-][-][-]]
97[[Zipf][-][X][-][-][-]]
98[[Birnbaum-Saunders / Fatigue Life][-][-][X][-][-]]
99[[Double Exponential][-][-][X][-][-]]
100[[Power Normal][-][-][X][-][-]]
101[[Power Lognormal][-][-][X][-][-]]
102[[Cosine][-][-][-][X][-]]
103[[Double Gamma][-][-][-][X][-]]
104[[Double Weibul][-][-][-][X][-]]
105[[Hyperbolic Secant][-][-][-][X][-]]
106[[Semicircular][-][-][-][X][-]]
107[[Bradford][-][-][-][X][-]]
108[[Birr / Fisk][-][-][-][X][-]]
109[[Reciprocal][-][-][-][X][-]]
110
111[/0 votes but useful anyway?]
112[[Kolmogorov Distribution][-][-][-][-][-]]
113]
114
115Also asked for more than once:
116
117* Add support for interpolated distributions, possibly combine with numeric
118integration and differentiation.
119* Add support for bivariate and multivariate distributions: most especially the normal.
120* Add support for the log of the cdf and pdf:
121this is mainly a performance optimisation since we can avoid
122some special function calls for some distributions
123by returning the log of the result.
124
125[endsect] [/section:issues Known Issues, and Todo List]
126
127[/
128  Copyright 2006, 2010 John Maddock and Paul A. Bristow.
129  Distributed under the Boost Software License, Version 1.0.
130  (See accompanying file LICENSE_1_0.txt or copy at
131  http://www.boost.org/LICENSE_1_0.txt).
132]
133
134