-
Notifications
You must be signed in to change notification settings - Fork 0
/
DESCRIPTION
69 lines (69 loc) · 4.04 KB
/
DESCRIPTION
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
Package: eikosograms
Type: Package
Title: The Picture of Probability
Version: 0.1.1.9000
Date: 2019-07-19
Authors@R: c(person(given = "Wayne", family = "Oldford",
email = "[email protected]",
role = c("aut", "cre")),
person(given = "Erle", family = "Holgersen",
email = "[email protected]",
role = "aut"),
person(given = "Ben", family = "Lafreniere",
email = "[email protected]",
role = "aut"),
person(given = "Tianlu", family = "Zhu",
email = "[email protected]",
role = "aut")
)
Author:
Wayne Oldford [aut, cre],
Erle Holgersen [aut],
Ben Lafreniere [aut],
Tianlu Zhu [aut]
Maintainer: Wayne Oldford <[email protected]>
URL: https://github.com/rwoldford/eikosograms
Description: An eikosogram (ancient Greek for probability picture) divides the unit square
into rectangular regions whose areas, sides, and widths, represent various probabilities
associated with the values of one or more categorical variates.
Rectangle areas are joint probabilities, widths are always marginal (though possibly joint
margins, i.e. marginal joint distributions of two or more variates), and heights of rectangles
are always conditional probabilities.
Eikosograms embed the rules of probability and are useful for introducing elementary probability
theory, including axioms, marginal, conditional, and joint probabilities, and their
relationships (including Bayes theorem as a completely trivial consequence).
They are markedly superior to Venn diagrams for this purpose, especially in distinguishing
probabilistic independence, mutually exclusive events, coincident events, and associations.
They also are useful for identifying and understanding conditional independence structure.
As data analysis tools, eikosograms display categorical data in a manner similar
to Mosaic plots, especially when only two variates are involved (the only case in which
they are essentially identical, though eikosograms purposely disallow spacing between rectangles).
Unlike Mosaic plots, eikosograms do not alternate axes as each new categorical variate
(beyond two) is introduced.
Instead, only one categorical variate, designated the "response", presents on the vertical axis
and all others, designated the "conditioning" variates, appear on the horizontal.
In this way, conditional probability appears only as height and marginal probabilities as widths.
The eikosogram is therefore much better suited to a response model analysis (e.g. logistic model)
than is a Mosaic plot.
Mosaic plots are better suited to log-linear style modelling as in discrete multivariate analysis.
Of course, eikosograms are also suited to discrete multivariate analysis with each variate in turn
appearing as the response.
This makes it better suited than Mosaic plots to discrete graphical models based on conditional
independence graphs (i.e. "Bayesian Networks" or "BayesNets").
The eikosogram and its superiority to Venn diagrams in teaching probability is described in
W.H. Cherry and R.W. Oldford (2003) <https://math.uwaterloo.ca/~rwoldfor/papers/eikosograms/paper.pdf>,
its value in exploring conditional independence structure and relation to graphical and log-linear models
is described in R.W. Oldford (2003) <https://math.uwaterloo.ca/~rwoldfor/papers/eikosograms/independence/paper.pdf>,
and a number of problems, puzzles, and paradoxes that are easily explained with eikosograms are given in
R.W. Oldford (2003) <https://math.uwaterloo.ca/~rwoldfor/papers/eikosograms/examples/paper.pdf>.
License: GPL-3
Depends: R (>= 3.1.0)
Imports: plyr, grid
LazyLoad: yes
LazyData: true
BugReports: https://github.com/rwoldford/eikosograms/issues
Encoding: UTF-8
RoxygenNote: 6.1.1
NeedsCompilation: no
Suggests: knitr, rmarkdown, gridExtra
VignetteBuilder: knitr, rmarkdown