TPreg: Regression Modeling of Transition Probabilities in a...

In idmTPreg: Regression Model for Progressive Illness Death Data

Description

Fits a semi-parametric regression model to estimate the effects on transition probabilities in a (possibly non-Markov) progressive illness-death model for a sequence of time.

Usage

TPreg(formula, data, link, s = 0, t = NULL, R = 199, by = NULL, trans, ncores = NULL)

Arguments

formula	an object of class formula which specifies the covariates. For example formula = ~ age + sex.
data	a data.frame of iddata class or a data.frame in which other than covariates five variables; id, Zt, delta1, Tt, delta are included.
link	a link function for binomial family which are logit, probit and cauchit, (corresponding to logistic, normal and Cauchy CDFs respectively).
s	the current time for the transition probabilities; default is zero which reports the occupation probabilities.
t	the Future time for the transition probabilities; default is NULL which is the largest uncensored sojourn time in the initial state.
R	the number of bootstrap replicates. Default is 199.
by	number: increment of the sequence from time s to time t.The default is NULL which is \lfloor (\max({ Zt}) - \min({Zt}))/q_{0.01}({ Zt }) \rfloor , where q_{0.01}(.) is the sample quantile corresponding to 0.01 probability and \lfloor x\rfloor gives the largest integer less than or equal to x. A binomial regression at every byth time between s and t is performed. by=1 reports all binomial regression results for each jump time, corresponding to the specified transition(s), between s and t. By increasing by we skip some times. In order to save the time, for a relatively large dataset a relatively big by is recommended.
trans	the possible transition(s) for a progressive illness-death model. For trans argument there are five options: "11", "12", "13", "23", and "all".
ncores	the number of cores to use for parallel execution. Default is the number of CPU cores on the current host.

Value

TPreg returns an object of class TPreg. An object of class TPreg is a list containing at least the following components:

co	the list of: ‘transition’ the transition, ‘time’ the jump times, ‘coefficients’ the estimated effects , ‘SD’ standard errors, ‘LWL’ lower confidence limits, ‘UPL’ upper confidence limits, ‘p.value’ p-values.
call	the matched call.
transition	the transition, this is equal to the transition in co unless for trans="all"
s	the current time for the transition probability.
t	the future time for the transition probability.
n.misobs	the number of missing observations.

In addition, trans="all" will have four lists: co11, co12, co13, and co23 instead of co and will give the information for all possible transitions in the progressive illness death model

References

Azarang, L. Scheike, TH. and de Una-Alvarez, J. (2017) Direct modeling of regression effects for transition probabilities in the progressive illness-death model, Statistics in Medicine 36, 1964-1976.

See Also

print.TPreg, summary.TPreg, and plot.TPreg as generic functions.

Examples

data(colonTPreg)

co11 <- TPreg( ~ Age + Nodes + treatment, colonTPreg, link = "logit", s = 50, R = 19, t = 200,
trans = "11", ncores = 1)
co11
summary(co11)

Example output

Loading required package: survival
Loading required package: foreach
Loading required package: doParallel
Loading required package: iterators
Loading required package: parallel
Warning message:
In TPreg(~Age + Nodes + treatment, colonTPreg, link = "logit", s = 50,  :
  Nodes variable in 'data' has missing value(s), 
Call:
TPreg(formula = ~Age + Nodes + treatment, data = colonTPreg, 
    link = "logit", s = 50, t = 200, R = 19, trans = "11", ncores = 1)

Transition:
[1] "11"

(s,t):
[1]  50 200

Coefficients:
                     Estimate     St.Err        LW.L        UP.L      P.value
X.Intercept.      2.492609419 0.72041023  1.08060538  3.90461346 0.0005402031
Age              -0.005980971 0.01181244 -0.02913336  0.01717142 0.6126264027
Nodes            -0.074748098 0.02923819 -0.13205494 -0.01744125 0.0105724084
treatmentLev      0.039682347 0.24418653 -0.43892324  0.51828794 0.8709055616
treatmentLev.5FU  0.929566117 0.32475154  0.29305310  1.56607913 0.0042045690


[1] "18 observations deleted due to missingness from 'data'"
Call:
TPreg(formula = ~Age + Nodes + treatment, data = colonTPreg, 
    link = "logit", s = 50, t = 200, R = 19, trans = "11", ncores = 1)
(s,t):
[1]  50 200

 Transition 11  :

     Coefficients:
  time X.Intercept.          Age       Nodes treatmentLev treatmentLev.5FU
1   56    58.363354 -0.401888881  6.26264521 -31.10806842       -9.4653733
2  185     2.636107 -0.007185251 -0.07924264   0.10143607        1.0470649
3  200     2.492609 -0.005980971 -0.07474810   0.03968235        0.9295661

     Standard Errors:
  time X.Intercept.        Age       Nodes treatmentLev treatmentLev.5FU
1   56  491.3913682 4.40435468 16.67490873  370.5809571      368.6372256
2  185    0.6884838 0.01042247  0.02954041    0.2612095        0.4063520
3  200    0.7204102 0.01181244  0.02923819    0.2441865        0.3247515

     Lower limit:
  time X.Intercept.         Age       Nodes treatmentLev treatmentLev.5FU
1   56  -904.763728 -9.03442405 -26.4201759 -757.4467443     -731.9943355
2  185     1.286679 -0.02761330  -0.1371418   -0.4105346        0.2506150
3  200     1.080605 -0.02913336  -0.1320549   -0.4389232        0.2930531

     Upper limit:
  time X.Intercept.        Age       Nodes treatmentLev treatmentLev.5FU
1   56  1021.490435 8.23064628 38.94546631  695.2306074       713.063589
2  185     3.985535 0.01324280 -0.02134344    0.6134067         1.843515
3  200     3.904613 0.01717142 -0.01744125    0.5182879         1.566079

     p.value:
  time X.Intercept.       Age       Nodes treatmentLev treatmentLev.5FU
1   56 0.9054562857 0.9272955 0.707234379    0.9331009      0.979515239
2  185 0.0001287392 0.4905717 0.007307052    0.6977702      0.009973593
3  200 0.0005402031 0.6126264 0.010572408    0.8709056      0.004204569


[1] "18 observation(s) deleted due to missingness from 'data'"

idmTPreg documentation built on May 2, 2019, 3:35 p.m.