discreteresponse: Copula models for count response data
In YafeiXu/CopulaModel: CopulaModel: Dependence Modeling with Copulas

Description Usage Arguments Value See Also Examples

Copula models for count response data, including discretized MVN diagnostics from bivariate margins

latentBVNnllk1(rho,param,ucdf,xdat1,xdat2,y1,y2)
latentBVNnllk2(rho,par1,par2,cdf1,cdf2,xdat1,xdat2,y1,y2)
ieenllk(param,upmf,ydat,xdat,LB,UB) # independent estimating equations
MVNlatent1(ydat,xdat,nrep,upmf,ucdf,upmfcdf,mx,ustart,LB,UB,prlevel=0)
   # discretized multivariate normal/Gaussian with univariate marginal
   # regression model being specified by upmf/ucdf and common
   # regression parameters for each margin
MVNlatent2(ydat,xdat,nrep,unllks,upmfs,ucdfs,upmfcdfs,mx,ustart,LB,UB,prlevel=0)
   # discretized multivariate normal/Gaussian with univariate marginal
   # regression models being specified by upmfs/ucdfs and different
   # regression parameters for each margin
vinebivEvsO1(ydat,xdat,nrep,mx,uparam,upmfcdf,cpar,A,pcop)

`rho`	latent correlation parameter
`param`	common parameter vector for univariate margin of the latent bivariate normal model; assumed estimated in a previous step in latentBVNnllk1
`par1`	univariate parameter vector for univariate margin 1 of the latent bivariate normal model; assumed estimated in a previous step
`par2`	univariate parameter vector for univariate margin 2 of the latent bivariate normal model; assumed estimated in a previous step
`uparam`	univariate parameter vector, common to all univariate margin
`ucdf`	function for univariate cdf
`upmf`	function for common univariate pmf for ieenllk
`upmfcdf`	function that efficiently computes univariate pmf and cdf up to an upper limit mx
`unllks`	vector of strings for univariate negative log-likelihoods
`upmfs`	vector of strings for pmf functions for margins 1,...,d
`ucdfs`	vector of strings for cdf functions for margins 1,...,d
`upmfcdfs`	vector of strings for pmf/cdf functions for computing up to mx
`mx`	bound used for Expected vs Observed tables in univariate/bivariate margins
`cdf1`	function for univariate cdf for margin 1
`cdf2`	function for univariate cdf for margin 1
`xdat1`	covariates for margin 1 (nxq), q=#covariates
`xdat2`	covariates for margin 2 (nxq), q=#covariates
`xdat`	covariate matrix ((n*d)xq), q=#covariates, d=#repeated measurements
`y1`	count response 1, vector of length n
`y2`	count response 2, vector of length n
`ydat`	count response, vector of length n*d
`nrep`	#repeated measurements per subject
`ustart`	starting parameter point for univariate model
`LB`	lower bound for param
`UB`	upper bound for param
`A`	dxd vine array with 1:d on diagonal
`pcop`	function for copula cdf (with scalar parameter) in tree 1 of vine
`cpar`	parameter values for pair-copulas in tree 1 of vine
`prlevel`	print.level for nlm()

negative log-likelihood value for latentBVNnllk1, latentBVNnllk2;

negative of sum of univariate log-likelihoods for ieenllk;

list(uparam vector or parmat matrix,rhvec,E1arr,O1arr,E2arr,O2arr) [univariate parameter estimates. latent correlation estimates, Expected and Observed counts for univariate/bivariate] for MVNlatent1,MVNlatent2;

list(rhosub,E1arr,O1arr,E2arr,O2arr) for mvnbivEvsO1 where rhosub is based on subset and rhopar is based on full data set;

list(E2arr,O2arr) for vinebivEvsO1 with Expected and Observed counts for bivariate margins in first tree of vine.

rvinediscrete

# count regression model with common betas for all margins: longitudinal count
data(rwmsubset)
rwm=rwmsubset
rwm$agec=(rwm$age-50)/10
rwm$ageq=(rwm$agec)^2
rwm$handfra=rwm$handper/100
xdat=cbind(rwm$sex,rwm$agec,rwm$ageq,rwm$hsat,rwm$handfra,rwm$univ)
xdat=as.matrix(xdat)
ydat=rwm$docvis
nc=ncol(xdat)
nc1=nc+1
out=MVNlatent1(ydat,xdat,nrep=5,upmf=nb1pmf,ucdf=nb1cdf,upmfcdf=nb1pmfcdf,
  mx=7,ustart=c(1.7,.3,.2,.1,-.2,.7,-.5,1.5),
  LB=c(rep(-20,nc1),0),UB=rep(10,nc1+1),prlevel=1)
print(out$uparam)
print(out$rhvec)
# 
# GP1 count regression models for each margin: uncommon regression coefficients 
data(kzrepmeas)
kz=kzrepmeas
kz$agehun=kz$age/100
xdat=cbind(kz$agehun,kz$sex,kz$msmok)
ydat=kz[,6]
nrep=4
outgp1=MVNlatent2(ydat,xdat,nrep,unllks=rep("gp1nllk",4),ucdfs=rep("gp1cdf",4),
  upmfcdfs=rep("gp1pmfcdf",4),
  mx=3,ustart=c(0,0,0,0,1),LB=c(-20,-20,-20,-20,0),UB=rep(10,5),prlevel=0)
print(outgp1$uparam)
print(outgp1$rhvec)