analyze.pvs: Analyze P-Values
In pvclass: P-Values for Classification

Description Usage Arguments Details Value Author(s) References See Also Examples

Graphical displays and quantitative analyses of a matrix of p-values.

1	analyze.pvs(pv, Y = NULL, alpha = 0.05, roc = TRUE, pvplot = TRUE, cex = 1)

`pv`	matrix with p-values, e.g. output of `cvpvs` or `pvs`.
`Y`	optional. Vector indicating the classes which the observations belong to.
`alpha`	test level, i.e. 1 - confidence level.
`roc`	logical. If `TRUE` and `Y` is not `NULL`, ROC curves are plotted.
`pvplot`	logical. If `TRUE` or `Y` is `NULL`, the p-values are displayed graphically.
`cex`	A numerical value giving the amount by which plotting text should be magnified relative to the default.

Displays the p-values graphically, i.e. it plots for each p-value a rectangle. The area of this rectangle is proportional to the the p-value. The rectangle is drawn blue if the p-value is greater than alpha and red otherwise.
If Y is not NULL, i.e. the class memberships of the observations are known (e.g. cross-validated p-values), then additionally it plots the empirical ROC curves and prints some empirical conditional inclusion probabilities I(b,θ) and/or pattern probabilities P(b,S). Precisely, I(b,θ) is the proportion of training observations of class b whose p-value for class θ is greater than α, while P(b,S) is the proportion of training observations of class b such that the (1 - α)-prediction region equals S.

`T`	Table containing empirical conditional inclusion and/or pattern probabilities for each class b. In case of L = 2 or L=3 classes, all patterns S are considered. In case of L > 3, all inclusion probabilities and some special patters S are considered.

Niki Zumbrunnen niki.zumbrunnen@gmail.com
Lutz Dümbgen lutz.duembgen@stat.unibe.ch
www.imsv.unibe.ch/duembgen/index_ger.html

Zumbrunnen N. and Dümbgen L. (2017) pvclass: An R Package for p Values for Classification. Journal of Statistical Software 78(4), 1–19. doi:10.18637/jss.v078.i04

Dümbgen L., Igl B.-W. and Munk A. (2008) P-Values for Classification. Electronic Journal of Statistics 2, 468–493, available at http://dx.doi.org/10.1214/08-EJS245.

Zumbrunnen N. (2014) P-Values for Classification – Computational Aspects and Asymptotics. Ph.D. thesis, University of Bern, available at http://boris.unibe.ch/id/eprint/53585.

cvpvs, pvs

X <- iris[c(1:49, 51:99, 101:149), 1:4]
Y <- iris[c(1:49, 51:99, 101:149), 5]
NewX <- iris[c(50, 100, 150), 1:4]

cv <- cvpvs(X,Y)
analyze.pvs(cv,Y)

pv <- pvs(NewX, X, Y, method = 'k', k = 10)
analyze.pvs(pv)

            
b               P(b,{})  P(b,{1})   P(b,{2})   P(b,{3}) P(b,{1,2}) P(b,{1,3})
  setosa     0.04081633 0.9591837 0.00000000 0.00000000          0          0
  versicolor 0.00000000 0.0000000 0.95918367 0.04081633          0          0
  virginica  0.00000000 0.0000000 0.04081633 0.91836735          0          0
            
b            P(b,{2,3}) P(b,{1,2,3})
  setosa     0.00000000            0
  versicolor 0.00000000            0
  virginica  0.04081633            0