collett (version 0.1.2)

step_s: Utilities

Description

Given a data-frame with an "s" step function, expand the data-frame to include the steps.

Given a survfit object, return a data-frame

Given a summary.survfit object, return a data-frame

Calculates lp=tt(x). Typically, the tt function should include an intercept term (see the examples below). Note that spline terms assume that the x argument is multiplicative; moreover, the additional arguments are not passed. For other types of tt terms, the x is passed directly to the tt function together with other arguments.

Usage

step_s(
  data,
  x,
  y,
  ymin,
  ymax,
  group,
  add_origin = TRUE,
  x_origin = 0,
  y_origin = 1
)
# S3 method for survfit
as.data.frame(x, row.names, optional, type = c("expanded", "plain"), ...)
# S3 method for summary.survfit
as.data.frame(x, row.names, optional, type = c("expanded", "plain"), ...)
predict_coxph_tt(object, times, type = "lp", se.fit = FALSE, x = 1, ...)
predict_coxph_tv(object, data, id)
plot_coxph_functional(
  formula,
  data,
  x = NULL,
  pch = 19,
  ylab = "Martingale residual for null model",
  smoother = c("loess", "lm"),
  smoother.formula = resi ~ xi,
  smoother.args = list(),
  points.args = list(),
  ...
)

Value

expanded data-frame with the same names

a vector of fitted values (when se.fit=FALSE) or a data-frame with fitted and se.fit columns (when se.fit=TRUE)

an update of data with survival probabilities

invisible plot return

Arguments

data

a dataset for evaluation of the coxph model

x

a numeric vector for the smoother (defaults to the 401 values between the range)

y

name of the y-variable (required)

ymin

name of the ymin variable (optional)

ymax

name of the ymax variable (optional)

group

name of a grouping variable (optional)

add_origin

logical for whether to add an origin to the start of each group

x_origin

double for the value of x at the origin

y_origin

double for the value of y, ymin and ymax at the origin

row.names

not used (in generic signature)

optional

not used (in generic signature)

type

a character for the type of prediction (currently only the linear predictor for the tt argument)

...

other arguments to pass to the plot function

object

a coxph object

times

a numeric vector of times to evaluate the linear predictor

se.fit

a logical for whether to return the standard errors

id

a character for the subject id

formula

a formula with a Surv on the lhs and a single variable on the rhs

pch

an integer for the pch argument in the plot for the residuals

ylab

a character for the ylab argument in the plot

smoother

a character for the name of the smoother

smoother.formula

a formula for the smoother in terms of resi and xi

smoother.args

a list of arguments to pass to the smoother function

points.args

a list of arguments to pass to the points function

Examples

  step_s(data.frame(g=c(1,1), a=1:2, b=4:5), a, b)
  step_s(data.frame(g=c(2,2), a=3:4, b=6:7), a, b)
  step_s(data.frame(g=c(1,1,2,2), a=1:4, b=4:7), a, b, group=g)
  library(survival)
  library(tinyplot)
  sfit1 = survfit(Surv(time,status)~rx, data=survival::colon,
                  subset=etype==1)
  with(as.data.frame(sfit1),
       tinyplot::plt(surv~time|strata,ymin=lower,ymax=upper,type="ribbon"))
  library(survival)
  library(tinyplot)
  sfit1 = survfit(Surv(time,status)~rx, data=survival::colon,
                  subset=etype==1)
  with(as.data.frame(sfit1, type="expanded"),
       tinyplot::plt(surv~time|strata,ymin=lower,ymax=upper,type="ribbon"))
 library(splines)
 library(tinyplot)
 fit1 = coxph(Surv(time,status)~tt(treat),data=breast_rfs,
              tt=function(x,t,...) x*cbind(1,t))
 fit2 = coxph(Surv(time,status)~tt(treat),data=breast_rfs,
              tt=function(x,t,...) x*ns(t,df=4,intercept=TRUE))
 times = seq(0,2500,len=301L)
 df1 = transform(predict_coxph_tt(fit1,times,se.fit=TRUE),
                 lower=exp(fitted-1.96*se.fit),
                 upper=exp(fitted+1.96*se.fit),
                 fitted=exp(fitted),
                 model="linear",times=times)
 df2 = transform(predict_coxph_tt(fit2,times,se.fit=TRUE),
                 lower=exp(fitted-1.96*se.fit),
                 upper=exp(fitted+1.96*se.fit),
                 fitted=exp(fitted),
                 model="ns",times=times)
 with(rbind(df1,df2),
      plt(fitted ~ times | model, ymin=lower, ymax=upper, type="ribbon",
          xlab="Time since diagnosis (days)",
          ylab="Hazard ratio comparing treated with untreated"))
 with(subset(breast_rfs,status==1), rug(time))
library(survival)
liver = transform(collett::liverbase, lbr=NULL)
liver = tmerge(liver, liver, id=patient, status=event(time,status))
liver = tmerge(liver,
               rbind(with(collett::liverbase, data.frame(patient,tstart=0,lbr)),
                     with(collett::lbrdata0, data.frame(patient,tstart=time,lbr))),
           id=patient, lbr = tdc(tstart,lbr))
fit3 = coxph(Surv(tstart,tstop,status)~lbr+treat,liver)
predict_coxph_tv(fit3,data=subset(liver,patient %in% c(1,7)),
                 id="patient")
library(survival)
par(mfrow=c(2,2))
plot_coxph_functional(Surv(time,status)~hb, data=collett::myeloma,
                      xlab="Value of Hb")
plot_coxph_functional(Surv(time,status)~bun, data=collett::myeloma,
                      xlab="Value of Bun")
plot_coxph_functional(Surv(time,status)~log(bun), data=collett::myeloma,
                      xlab="Value of log Bun")