| p_value.pois {Fuzzy.p.value} | R Documentation |
Testing hypotheses based on fuzzy p-value for Poisson distribution
Description
Function p_value.pois can draw the membership function of fuzzy p-value for the following three major problems which can be usually considered for the following tests in a fuzzy environment:
(1) testing crisp hypotheses based on fuzzy data,
(2) testing fuzzy hypotheses based on crisp data, and
(3) testing fuzzy hypotheses based on fuzzy data.
Also, one can consider a fuzzy significance of level for each of above tests by function p_value.pois.
It must be noted that function p_value.pois is applied when the test statistics has Poisson distribution. This idea can be extended for other test statistics which has other discrete distribution.
All needed formulas are presented in Springer Chapter book 2016 and the considered examples are match with this reference for the readers.
Usage
p_value.pois(kind, H0, H1, t, n, sig)
Arguments
kind |
The kind of testing hypotheses is one of three numbers 0, 1 and 2, which is distinguishable from the form of alternative (fuzzy/non-fuzzy) hypothesis.
Set |
H0 |
The null hypothesis of the test which must introduce only via functions |
H1 |
The alternative hypothesis of the test which must introduce only via functions |
t |
The observed value of the test statistic (the observed value of the mean), which must introduce only via functions |
n |
Sample size |
sig |
The significance of level for the test which must introduce only via functions |
Details
In this package, the users must introduce H0, H1, t and significance level only by fuzzy numbers T, Tr, S or B. Therefore, if each of them are precise (crisp), user must introduce it by special fuzzy sets (i.e. by indicator functions).
For example the crisp null hypothesis "H_0: \ \lambda > 5.2" must be introduced by H0 = B(5.2, 5.2) to function p_value.pois.
Value
The function p_value.pois is constructed for hypotheses testing in fuzzy environment where the test statistics has Poisson distributions and the parameter which discussed by two hypotheses is the mean of Poisson.
The major duty of function p_value.pois is:
(1) draw the membership functions of p-value and significance level for a Poisson test,
(2) compute numerical measures of \Delta_{SP} and \Delta_{PS} (see Parchami and Mashinchi (2016) for formulas),
(3) decide to accept null hypothesis or accept alternative hypothesis in the test, and finally
(4) compute the degree of acceptance for the accepted hypothesis for a Poisson test.
Author(s)
Abbas Parchami
References
Filzmoser, P., and Viertl, R. (2004). Testing hypotheses with fuzzy data: the fuzzy p-value. Metrika 59: 21-29.
Holena, M. (2004). Fuzzy hypotheses testing in a framework of fuzzy logic. Fuzzy Sets and Systems 145: 229-252.
Parchami, A., Taheri, S. M., and Mashinchi, M. (2010). Fuzzy p-value in testing fuzzy hypotheses with crisp data. Statistical Papers 51: 209-226.
Parchami, A., Taheri, S. M., and Mashinchi, M. (2012). Testing fuzzy hypotheses based on vague observations: a p-value approach. Statistical Papers 53: 469-484.
Wang, X., Kerre, E. E. (2001). Reasonable properties for the ordering of fuzzy quantities (II). Fuzzy Sets and Systems 118: 387-405.
Yuan, Y. (1991). Criteria for evaluating fuzzy ranking methods. Fuzzy Sets Syst 43: 139-157.
Examples
# Solving Example 5 from Springer Chapter book 2016
# Considering the assumptions of this example, it must be note that H0 and H1 are not about
# the parameter of the test statistics, and in fact hypotheses must be 12*H0 and 12*H1
H0 = B( 12*2.75 , 12*3.25 )
H1 = S( 12*2.75 , 12*3.25 )
t = T( 27, 27, 27)
sig = T(0.0,0.05,0.1)
p_value.pois( kind=1, H0, H1, t, n=12, sig)
# Solving Example 7 from Springer Chapter book 2016
H0 = B( 12*2.75 , 12*3.25 )
H1 = S( 12*2.75 , 12*3.25 )
t = T( 24, 27, 30)
sig = T(0.0,0.05,0.1)
p_value.pois( kind=1, H0, H1, t, n=12, sig)
## The function is currently defined as
function (kind, H0, H1, t, n, sig)
{
alpha_L = seq(0, 1, 0.01)
alpha_U = seq(1, 0, -0.01)
if (kind == 1) {
p_L = ppois(alphacut(t, alpha_L)[, "U"], alphacut(H0,
alpha_L)[, "L"])
p_U = ppois(alphacut(t, alpha_U)[, "L"], alphacut(H0,
alpha_U)[, "U"])
}
else if (kind == 2) {
p_L = 1 - ppois(alphacut(t, alpha_L)[, "U"], alphacut(H0,
alpha_L)[, "L"])
p_U = 1 - ppois(alphacut(t, alpha_U)[, "L"], alphacut(H0,
alpha_U)[, "U"])
}
else if (supp(t)[1] >= supp(H0)[2]) {
p_L = 2 * (1 - ppois(alphacut(t, alpha_L)[, "U"], alphacut(H0,
alpha_L)[, "L"]))
p_U = 2 * (1 - ppois(alphacut(t, alpha_U)[, "L"], alphacut(H0,
alpha_U)[, "U"]))
}
else if (supp(t)[2] <= supp(H0)[1]) {
p_L = 2 * ppois(alphacut(t, alpha_L)[, "U"], alphacut(H0,
alpha_L)[, "L"])
p_U = 2 * ppois(alphacut(t, alpha_U)[, "L"], alphacut(H0,
alpha_U)[, "U"])
}
else {
return("The fuzzy p-value can not defined for this example, since the fuzziness of
the problem is very high. This case may be accured for the testing equality
fuzzy null hypothesis, against the non-equality fuzzy althenative hypothesis")
}
alpha = cbind(alpha_L, alpha_U)
p = cbind(p_L, p_U)
plot(p, alpha, type = "l", lwd = 3, col = 1, xlim = c(-0.04,
1.02))
s1 = supp(sig)[1]
s2 = core(sig)[1]
s3 = core(sig)[1]
s4 = supp(sig)[2]
lines( c(0,s1,s2,s3,s4,1), c(0,0,1,1,0,0), type='l', lty=3, lwd=2, col=2 )
# plot(sig, lty=3, lwd=2, col=2, add=TRUE) # Conflict importing 2 plot from different packages
legend("topright", c("Fuzzy p-value", "Significance level"),
col = c(1, 2), text.col = 1, lwd = c(3, 2), lty = c(1,
3))
if( class(sig) == "numeric" ){
sig <- TriangularFuzzyNumber(sig, sig, sig)
}
P_L = p_L
P_U = p_U
knot.n = 100
S_L = alphacut(sig, round(seq(0, 1, .01), 5))[,"L"]
S_U = alphacut(sig, round(seq(0, 1, .01), 5))[,"U"]
Int1 = ( P_U - S_L ) * ( P_U > S_L )
Int2 = ( P_L - S_U ) * ( P_L > S_U )
Arz = 1 / (knot.n - 1) #Arze Mostatilha baraye mohasebe-ye Integral
Integral1 <- ( sum( Int1 ) - Int1[1]/2 - Int1[length(Int1)]/2 ) *Arz
Integral2 <- ( sum( Int2 ) - Int2[1]/2 - Int2[length(Int2)]/2 ) *Arz
Delta_PS = Integral1 + Integral2
Int3 = ( S_U - P_L ) * ( S_U > P_L )
Int4 = ( S_L - P_U ) * ( S_L > P_U )
Integral3 <- ( sum( Int3 ) - Int3[1]/2 - Int3[length(Int3)]/2 ) *Arz
Integral4 <- ( sum( Int4 ) - Int4[1]/2 - Int4[length(Int4)]/2 ) *Arz
Delta_SP = Integral3 + Integral4
print("Delta_SP = ")
print(Delta_SP)
print("Delta_PS = ")
print(Delta_PS)
Degree_P_biger_than_S = Delta_PS/(Delta_PS + Delta_SP)
Degree_S_biger_than_P = 1 - Degree_P_biger_than_S
if (Degree_P_biger_than_S >= Degree_S_biger_than_P) {
a = "The null hypothesis (H0) is accepted with degree D(P>S)="
b = round(Degree_P_biger_than_S, 4)
c = ", at the considered significance level."
noquote(sprintf("%s %s %s ", a, b, c))
}
else {
if (Degree_P_biger_than_S < Degree_S_biger_than_P) {
a = "The althernative hypothesis (H1) is accepted with degree D(S>P)="
b = round(Degree_S_biger_than_P, 4)
c = ", at the considered significance level."
noquote(sprintf("%s %s %s ", a, b, c))
}
else {
return(noquote(paste0("Impossible case")))
}
}
}