n8thangreen

##e.g. data from kmi in LOS_toy_example.R
##non-infected only
#data <- data[-173,]

## (death) count at each event time
nDx <- table(data$exit)

## event times
times <- as.numeric(names(table(data$exit)))

n8thangreen

library("R2wd")

require(R2wd)


wdGet(T)  # If no word file is open, it will start a new one - can set if to have the file visiable or not
wdNewDoc("H:\\This.doc")	# this creates a new file with "this.doc" name

wdApplyTemplate("H:\\This.dot")	# this applies a template

n8thangreen

charlson2 <- function (Patient, icd) 
{
  ##
  ##
  ##
  ## foo <- charlson2(Patient=survData$Spell, icd=survData$Prim.Diag.Code)
  
  d <- cbind.data.frame(Patient, icd)
  nobs <- nrow(d)
  d$code <- rep(NA, nobs)

n8thangreen

############################
###GUM2011AllPositivity#####
############################
summary(dat$totalPositivity2011GUM)
hist(dat$totalPositivity2011GUM)

modpos<-glm(totalPositivity2011GUM~ratioWhite2011+ratioBlack2011+ratioAsian2011+
	conception.rate.per.1000women.Under18+MarriedCivilCohabitCouples2011+SingleParents2011+
	ratioNoQual_2011+ratioMales25to34_2011+RatioFemales16to24_2011,
	family=binomial,data=dat,weights=sum.NtestsGUM2011All)

n8thangreen

#####################################
#
# Cox regression aggregated functions
#
# Nathan Green
# 11-2012
#
#####################################

n8thangreen

tvCox <- function(survData, covariates, type){
  ##
  ## time-dependent hazard array representation
  ## can specify which time to use: cause specific or subdistributions
  ##
  ## survData: survival data subset of total.data with indicators
  ## covariates: e.g. c("age", "agegr", "cath", "surgical", "gender")
  ## type: type of hazard functions to use
  ##         "" - Cause-specific
  ##		"death" - subdistribution time-to-death 

n8thangreen

subdistn <- function(survData){
  ## substitute in later times to change risk sets
  ## time representation appropriate for subdistribution hazards
  ## 
  ## Accounts for administrative censoring
  ## note that within a specific organism subset the maximum time
  ## may be from a patient infected by a different organism
  
  ## default entries

n8thangreen

multinom_loglik <- function(props, size.melt, cov){
  ##
  ## multinomial logistic regression likelihood
  ## over ALL groups in one batch 
  ##
  ## props: array of covariate by group probabilities
  ## size.melt: e.g. agesize.melt
  ## cov: covariate name (string) e.g. "age"
  
  ncol.size.melt <- ncol(size.melt)

n8thangreen

probs.logit <- function(grps.seq, cov, coeff, x){
  ##
  ## multinomial logistic regression model fit for a given covariate
  ## e.g. probability sums per age =1
  ##
  ## x: covariate value e.g. age=16
  ## grps: vector of class labels for NATSAL groups e.g. het5yrs
  ## cov: covariate name to regress against (string)
  ## coeff: coefficients from the mlogit() fit
  

n8thangreen

probs.logit_covs <- function(groups, cov, data){
  ##
  ## given a (set of) covariate values, calculates the class membership probabilities
  ##
  ## groups: response field name of interest in NATSAL e.g. het1yr (string)
  ## cov: explanatory covariate name e.g. age_shift (string)
  ## data: e.g. NATSAL.dat (array)
  
  ## ordered sequence of covariate levels
  ## copy those covered in data but could include intermediate values too