That's my latin hypercube sampling of an epidemiological model:
parameters <- c(Beta = 1.8, gamma_h= 1/7, alpha= 5, Beta2= 3, mu_m= 1/8, C= 1000, p= 0.6)
lhs<-maximinLHS(100,length(parameters))
Beta.min <- 0.1
Beta.max <- 6
gamma_h.min <- 1/10
gamma_h.max <- 1/4
alpha.min <- 3
alpha.max <- 7
Beta2.min <- 1
Beta2.max <- 6
mu_m.min <- 1/12
mu_m.max <- 1/6
C.min <- 500
C.max <- 2000
p.min <- 0.3
p.max= 0.9
LHS_SIRV.uniform <- cbind( Beta = lhs[,1]*(Beta.max-Beta.min)+Beta.min,
gamma_h = lhs[,2]*(gamma_h.max-gamma_h.min)+gamma_h.min,
alpha = lhs[,3]*(alpha.max-alpha.min)+alpha.min,
Beta2 = lhs[,4]*(Beta2.max-Beta2.min)+Beta2.min,
mu_m = lhs[,5]*(mu_m.max-mu_m.min)+mu_m.min,
C = lhs[,6]*(C.max-C.min)+C.min,
p = lhs[,7]*(p.max-p.min)+p.min)
Then after that I want to do a partial rank correlation coefficient. For that I create a dataframe of parameters and main output :
parameters <- c(Beta = 1.8, gamma_h= 1/7, alpha= 5, Beta2= 3, mu_m= 1/8, C= 1000, p= 0.6)
init <- c(S_h= 65000, I_h = 1, R_h=0, D_h= 0, S_m= 30000, I_m= 1)
SIRV <- data.frame(matrix(rep(NA,100),nrow=100, ncol= ncol(LHS_SIRV.uniform)+3))
colnames(SIRV)<-c(names(parameters), "MaxInf", "MaxDur", "Thresh1")
Then when I do the prcc procedure:
prcc_size <- pcc(SIRV[,1:length(parameters)], SIRV$MaxInf, nboot = 500, rank=TRUE, conf=1-bonferroni.alpha)
I ve the return :
"[1] "All values of t are equal to 1 \n Cannot calculate confidence intervals"
There were 50 or more warnings (use warnings() to see the first 50".
Call:
pcc(X = SIRV[, 1:length(parameters)], y = SIRV$MaxInf, rank = TRUE, nboot = 500, conf =
1 - bonferroni.alpha)
Partial Rank Correlation Coefficients (PRCC):
original bias std. error min. c.i. max. c.i.
Beta NA NA NA NA NA
gamma_h 0.66456921 NA NA 0.3298834 2.3285923
alpha -0.73157883 NA NA -2.4631431 -0.4668000
Beta2 -0.01338792 NA NA -1.0265995 0.9730587
mu_m -0.72973840 NA NA -2.4591508 -0.4602121
C NA NA NA NA NA
p 1.00000000 0 8.375662e-17 NA NA
Why Beta and C are NA ?????
Moeover when i do the same thing but with prcc_duration <- pcc(SIRV[,1:length(parameters)], SIRV$Thresh1, nboot = 500, rank=TRUE, conf=1-bonferroni.alpha)
I ve no problem :
Partial Rank Correlation Coefficients (PRCC):
original bias std. error min. c.i. max. c.i.
Beta -0.41777323 0.28176673 0.7894458 -1.83554646 0.1644535
gamma_h 0.51049624 -0.14188507 0.7348207 0.02099249 2.0209925
alpha 0.59870283 -0.06944773 0.6952997 0.19740566 2.1974057
Beta2 -0.11555968 0.10389373 0.7790595 -1.23111937 0.7688806
mu_m 0.65873910 -0.24657407 0.7308407 0.31747820 2.3174782
C -0.06936082 0.19371286 0.7691380 -1.13872163 0.8612784
p -0.34604591 0.10394436 0.7436426 -1.69209181 0.3079082