downward: performs the downward step of the peeling algorithm and...

In abureau/LCAextend: Latent Class Analysis (LCA) with familial dependence in extended pedigrees

Description

computes the probability of measurements above connectors and their classes given the model parameters, and returns the unnormalized triplet and individual weights. This is an internal function not meant to be called by the user.

Usage

downward(id, dad, mom, status, probs, fyc, peel, res.upward)

Arguments

id	individual ID of the pedigree,
dad	dad ID,
mom	mom ID,
status	symptom status: (2: symptomatic, 1: without symptoms, 0: missing),
probs	a list of probability parameters of the model,
fyc	a matrix of n times K+1 given the density of observations of each individual if allocated to class k, where n is the number of individuals and K is the total number of latent classes in the model,
peel	a list of pedigree peeling containing connectors by peeling order and couples of parents,
res.upward	result of the upward step of the peeling algorithm, see upward.

Details

This function computes the probability of observations above connectors and their classes using the function downward.connect, for each connector, if Y_above(i) is the observations above connector i and S_i and C_i are his status and his class respectively, the functions computes P(Y_above(i),S_i,C_i) by computing a downward step for the parent of connector i who is also a connector. These quantities are used by the function weight.nuc to compute the unnormalized triplet weights ww and the unnormalized individual weights w.

Value

The function returns a list of 2 elements:

ww	unnormalized triplet weights, an array of n times 2 times K+1 times K+1 times K+1, where n is the number of individulas and K is the total number of latent classes in the model, see e.step for more details,
w	unnormalized individual weights, an array of n times 2 times K+1, see e.step.

References

TAYEB et al.: Solving Genetic Heterogeneity in Extended Families by Identifying Sub-types of Complex Diseases. Computational Statistics, 2011, DOI: 10.1007/s00180-010-0224-2.

See Also

See also downward.connect.

Examples

#data
data(ped.cont)
data(peel)
fam <- ped.cont[,1]
id <- ped.cont[fam==1,2]
dad <- ped.cont[fam==1,3]
mom <- ped.cont[fam==1,4]
status <- ped.cont[fam==1,6]
y <- ped.cont[fam==1,7:ncol(ped.cont)]
peel <- peel[[1]]
#standardize id to be 1, 2, 3, ...
id.origin <- id
standard <- function(vec) ifelse(vec%in%id.origin,which(id.origin==vec),0)
id <- apply(t(id),2,standard)
dad <- apply(t(dad),2,standard)
mom <- apply(t(mom),2,standard)
peel$couple <- cbind(apply(t(peel$couple[,1]),2,standard),
                     apply(t(peel$couple[,2]),2,standard))
for(generat in 1:peel$generation) 
    peel$peel.connect[generat,] <- apply(t(peel$peel.connect[generat,]),2,standard)
#probs and param
data(probs)
data(param.cont)
#densities of the observations
fyc <- matrix(1,nrow=length(id),ncol=length(probs$p)+1)
fyc[status==2,1:length(probs$p)] <- t(apply(y[status==2,],1,dens.norm,param.cont,NULL))
#the upward step
res.upward <- upward(id,dad,mom,status,probs,fyc,peel)
#the function
downward(id,dad,mom,status,probs,fyc,peel,res.upward)

abureau/LCAextend documentation built on May 3, 2019, 9:41 p.m.