fitLM: Fit a coxph or CSC model to a LM super dataset

In anyafries/dynLM: Dynamic w-year risk predictions from landmark time points

Fit a coxph or CSC model to a LM super dataset

Description

Fit a coxph or CSC model to a LM super dataset

Usage

fitLM(
  formula,
  LMdata,
  type = "coxph",
  method = "breslow",
  func_covars,
  func_LMs,
  LM_col,
  outcome,
  w,
  LMcovars,
  ...
)

Arguments

formula	The formula to be used, remember to include "+cluster(ID)" for the column that indicates the ID of the individual for robust error estimates.
LMdata	An object of class "LMdataframe", this can be created by running cutLMsuper and addLMtime
type	"coxph" or "CSC"/"CauseSpecificCox"
method	A character string specifying the method for tie handling. Default is "breslow". More information can be found in coxph.
func_covars	A list of functions to use for interactions between LMs and covariates.
func_LMs	A list of functions to use for transformations of the landmark times.
LM_col	Character string specifying the column name that indicates the landmark time point for a row.
outcome	List with items time and status, containing character strings identifying the names of time and status variables, respectively, of the survival outcome
w	Scalar, the value of the prediction window (ie predict w-year/other time period risk from the LM points)
LMcovars	Vector of strings indicating the columns that are to have a LM interaction
...	Arguments given to coxph or CSC.

Value

An object of class "LMcoxph" or "LMCSC" with components:

• model: fitted model

• type: as input

• w, func_covars, func_LMs, LMcovars, allLMcovars, outcome: as in LMdata

• LHS: the LHS of the input formula

• linear.predictors: the vector of linear predictors, one per subject. Note that this vector has not been centered.

Examples

## Not run: 
data(relapse)
outcome = list(time="Time", status="event")
covars = list(fixed=c("ID","age.at.time.0","male","stage","bmi"),
              varying=c("treatment"))
w = 60; LMs = c(0,12,24)
# Covariate-landmark time interactions
func.covars <- list( function(t) t, function(t) t^2)
# let hazard depend on landmark time
func.LMs <- list( function(t) t, function(t) t^2)
# Choose covariates that will have time interaction
pred.covars <- c("age","male","stage","bmi","treatment")
# Stack landmark datasets
LMdata <- cutLMsuper(relapse, outcome, LMs, w, covs, format="long", id="ID", rtime="fup_time", right=F)
# Update complex LM-varying covariates
LMdata$LMdata$age <- LMdata$LMdata$age.at.time.0 + LMdata$LMdata$LM/12 # age is in years and LM is in months
# Add LM-time interactions
LMdata <- addLMtime(LMdata, pred.covars, func.covars, func.LMs)
formula <- "Hist(Time, event, LM) ~ age + male + stage + bmi + treatment + age_1 + age_2 + male_1 + male_2 + stage_1 + stage_2 + bmi_1 + bmi_2 + treatment_1 + treatment_2 + LM_1 + LM_2 + cluster(ID)"
supermodel <- fitLM(as.formula(formula), LMdata, "CSC")

## End(Not run)

anyafries/dynLM documentation built on July 26, 2022, 12:17 a.m.