JMcmprsk (version 0.9.10)

jmc: An integrated function for reconstructing data and do the joint modelling

Description

Reconstruct data into a regular longitudinal format as a refined dataset and do joint modelling for this refined data with continuous outcome.

Usage

jmc(
  long_data,
  surv_data,
  out,
  FE,
  RE,
  ID,
  cate = NULL,
  intcpt = 1,
  quad.points = 10,
  max.iter = 10000,
  quiet = TRUE,
  do.trace = FALSE
)

Arguments

long_data

Data matrix for longitudinal in long form. The time variable should be labeled 'time'.

surv_data

Data matrix for competing risks data. Each subject has one row of observation (as opposed to the long_data). First and second column should be the observed event time and censoring indicator, respectively. The coding for the censoring indicator is as follows: 0 - censored events, 1 - risk 1, 2 - risk 2. Two competing risks are assumed.

out

Column name for outcome variable in long_data.

FE

Vector of column names that correspond to the fixed effects in long_data. If missing, then all columns except for the outcome and ID columns will be considered.

RE

Types/Vector of random effects in long_data. The available type are "intercept", "linear", "quadratic" (time-related random effect specification) or other covariates in the input dataset. If specify other covariates, then they to be numerical vectors.

ID

Column name for subject ID number in long_data.

cate

Vector of categorical variables in long_data. Default is NULL.

intcpt

Specify either 0 or 1. Default is set as 1. 0 means no intercept in random effect.

quad.points

Number of quadrature points used in the EM procedure. Default is 20. Must be an even number. Larger values means higher accuracy but more time-consuming.

max.iter

Max iterations. Default is 10000.

quiet

Logical. Print progress of function. Default is TRUE.

do.trace

Logical. Print the parameter estimates during the iterations. Default is FALSE.

Value

Object of class JMcmprsk with elements

vcmatrix	The variance-covariance matrix for all the parameters. The parameters are in the order: \(\beta\), \(\sigma^2\), \(\gamma\), \(\nu\), and \(\Sigma\). The elements in \(\Sigma\) are output in the order along the main diagonal line, then the second main diagonal line, and so on.
betas	The point estimates of \(\beta\).
se_betas	The standard error estimate of \(\beta\).
gamma_matrix	The point estimate of \(\gamma\).
se_gamma_matrix	The standard error estimate of \(\gamma\).
v_estimate	The point estimate of \(\nu\).
se_v_estimate	The standard error estimate of \(\nu\).
sigma2_val	The point estimate of \(\sigma^2\).
se_sigma2_val	The standard error estimate of \(\sigma^2\).
sigma_matrix	The point estimate of \(\Sigma\) (only the upper triangle portion of the matrix is output).
se_sigma	The standard error estimate of \(\Sigma\).The standard errors are given in this order: main diagonal, the second main diagonal, and so on.
loglike	Log Likelihood.

See Also

jmc_0

Examples

# NOT RUN {
require(JMcmprsk)
set.seed(123)
data(lung)
yread <- lung[, c(1,2:11)]
cread <- unique(lung[, c(1, 12, 13, 6:10)])
# }
# NOT RUN {
#Please note only those variables that will appear in the model can be included
res <- jmc(long_data = yread, surv_data = cread, out = "FVC",
           FE = c("time", "FVC0", "FIB0", "CYC", "FVC0.CYC",
                  "FIB0.CYC", "time.CYC"),
           RE = "linear", ID = "ID",cate = NULL, intcpt = 0,
           quad.points = 8, quiet = FALSE)
# }
# NOT RUN {
#make up two categorical variables and add them into yread
sex <- sample(c("Female", "Male"), nrow(cread), replace = TRUE)
race <- sample(c("White", "Black", "Asian", "Hispanic"),
               nrow(cread), replace = TRUE)
ID <- cread$ID
cate_var <- data.frame(ID, sex, race)
if (require(dplyr)) {
  yread <- dplyr::left_join(yread, cate_var, by = "ID")
}
# }
# NOT RUN {
# run jmc function again for yread file with two added categorical variables
res2 <- jmc(long_data = yread, surv_data = cread,
            out = "FVC", cate = c("sex", "race"),
            FE = c("time", "FVC0", "FIB0", "CYC", "FVC0.CYC",
                "FIB0.CYC", "time.CYC"),
            RE = "time", ID = "ID", intcpt = 0,
            quad.points = 8, quiet = FALSE)
res2
# Extract the parameter estimates of longitudinal sub-model fixed effects
 beta <- coef(res2, coeff = "beta")
 beta
## Linear hypothesis of testing all coefficients of beta's equal 0
 linearTest(res2, coeff="beta")
 ## Linear hypothesis of testing beta1=beta2
 ## create a linear contrast for beta1=beta2 (intercept not included in Lb)
 Lb <- matrix(c(1, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0), ncol = length(beta)-1, nrow = 1)
 linearTest(res2, coeff="beta", Lb = Lb)
 # Extract the parameter estimates of survival sub-model fixed effects
 gamma <- coef(res2, coeff = "gamma")
 gamma
 ## Linear hypothesis of testing all coefficients of gamma's equal 0
 linearTest(res2, coeff="gamma")
 ## Linear hypothesis of testing gamma11=gamma21
 ##  (the coefficients of first covariate from
 ##    both risk functions are equal)
 Lg <- matrix(c(1, 0, 0, 0, 0, -1, 0, 0, 0, 0), ncol = length(gamma), nrow = 1)
 linearTest(res2, coeff="gamma", Lg = Lg)
 ## Extract the standard errors for the longitudinal portion
 summary(res2, coeff = "longitudinal", digits = 4)
 ## Extract the standard errors for the survival portion
 summary(res2, coeff = "survival", digits = 4)
 
# }