Title: | ccostr - Estimation of mean costs in censored data |
---|---|
Description: | This package implements various estimators for inferring the mean censored cost data. |
Authors: | Lars Boerty [aut, cre], Rasmus Broendum [aut], Martin Boegsted [aut] |
Maintainer: | Lars Boerty <[email protected]> |
License: | MIT + file LICENSE |
Version: | 0.1.0 |
Built: | 2024-11-24 04:09:54 UTC |
Source: | https://github.com/larshernandez/ccostr |
This function calculates the mean cost for right-censored cost data over a period of L time units (days, months, years,...)
ccmean(x, L = max(x$surv), addInterPol = 0)
ccmean(x, L = max(x$surv), addInterPol = 0)
x |
A dataframe with columns: id, cost, delta and surv. If Cost history is available it can be specified by: start and stop, |
L |
Limit. Mean cost is calculated up till L, if not specified L = max(surv) |
addInterPol |
This parameter affects the interpolation of cost between two observed times. Defaults to zero. |
The function returns four estimates. The first two are simple and biased downwards, and included for comparison. The estimates are:
- AS: "Available Sample estimator" - The simple sample mean
- CC: "Complete Case estimator" - The mean of fully observed cases
- BT: "Weighted Complete Case estimator" - Bang and Tsiatis's estimator
- ZT: "Weighted Available estimator" - Zhao and Tian's estimator
The function needs the following in a dataframe:
- id: The id separating each individual
- cost: The total cost, or if start and stop provided the specific cost
- start: Start of cost
- stop: End of cost, if one time cost then start = stop
- delta: Event variable, 1 = event, 0 = no event
- surv: Survival
An object of class "ccobject".
Bang H, Tsiatis AA (2000). “Estimating medical costs with censored data.” Biometrika, 87(2), 329–343. ISSN 00063444, doi:10.1093/biomet/87.2.329, http://www.ncbi.nlm.nih.gov/pubmed/304515374.
Zhao H, Tian L (2001). “On Estimating Medical Cost and Incremental Cost-Effectiveness Ratios with Censored Data.” Biometrics, 57(4), 1002–1008. ISSN 0006341X, doi:10.1111/j.0006-341X.2001.01002.x, http://www.ncbi.nlm.nih.gov/pubmed/11764238.
hcost ccmean(hcost, L = 1461, addInterPol = 1)
hcost ccmean(hcost, L = 1461, addInterPol = 1)
Simulated data from the stata hcost package
data(hcost)
data(hcost)
An object of Rdata
Chen S, Rolfes J, Zhao H (2015). “Estimation of Mean Health Care Costs and Incremental Cost-effectiveness Ratios with Possibly Censored Data.” The Stata Journal: Promoting communications on statistics and Stata, 15(3), 698–711. ISSN 1536-867X, doi:10.1177/1536867X1501500305, The Stata Journal, http://www.ncbi.nlm.nih.gov/pubmed/25080530.
data(hcost)
data(hcost)
Adding to the generic plot function
## S3 method for class 'ccobject' plot(x, ...)
## S3 method for class 'ccobject' plot(x, ...)
x |
The ccobject |
... |
passthrough |
a plot
Adding to the generic print function
## S3 method for class 'ccobject' print(x, ...)
## S3 method for class 'ccobject' print(x, ...)
x |
The ccobject |
... |
passthrough |
a plot
This function can be used to demonstrate the bias and coverage of the estimators in the ccmean function
simCostData(n = 100, dist = "unif", censor = "light", cdist = "exp", L = 10)
simCostData(n = 100, dist = "unif", censor = "light", cdist = "exp", L = 10)
n |
Number of individuals to simulate |
dist |
Survival distribution either "unif" = unif(0,10) o r "exp" = exp (1/6) |
censor |
Censoring "light" ~ 25% or "heavy" ~ 40%, changes a bit depending on cdist |
cdist |
Distribution used to censor, "exp" exponential or "unif" uniform |
L |
Number of years to summarize over |
The function simulates survival times from either an uniform distribution or an exponential distribution, and a cost history. There are two options for censoring, heavy (~40 light (~25
Simulation of censored cost
Lin DY, Feuer EJ, Etzioni R, Wax Y (1997). “Estimating Medical Costs from Incomplete Follow-Up Data.” Biometrics, 53(2), 419. ISSN 0006341X, doi:10.2307/2533947, http://www.ncbi.nlm.nih.gov/pubmed/9192444.
# The simulated data can be used to show how the estimators perform simCostData(n = 100, dist = "unif", censor = "light", cdist = "exp", L = 10)
# The simulated data can be used to show how the estimators perform simCostData(n = 100, dist = "unif", censor = "light", cdist = "exp", L = 10)