Description Usage Arguments Details Value Examples
Fits the linear ERR model on matched case-control data and performs first and second order jackknife correction
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 |
data |
data frame containing matched case-control data, with a number of columns for doses to different locations, a column containing matched set numbers, a column containing the case's tumor location (value between 1 and the number of locations, with location x corresponding to the x-th column index in |
doses |
vector containing the indices of columns containing dose information. |
set |
column index containing matched set numbers. |
status |
column index containing case status. |
loc |
column index containing the location of the matched set's case's second tumor. |
corrvars |
vector containing the indices of columns containing variables to be corrected for. |
ccmethod |
choice of method of analysis: one of meandose, CCML, CCAL or CL. Defaults to CCAL |
repar |
reparametrize to β=exp(ξ)? Defaults to |
initpars |
initial values for parameters, default is 0 for all parameters. If supplying a different vector, use a vector with an initial value for β or ξ, one for all of the other location effects and one for each other covariate (in that order). Note that if |
fitopt |
list with options to pass to |
uplimBeta |
upper limit for β=exp(ξ), default value 5. This is used for constraining the MLE estimation in some settings and for the jackknife inclusion criteria, and can be infinite except when Brent optimization is used (see details) |
profCI |
boolean: compute 95% profile likelihood confidence interval for β/ξ? Default value TRUE. |
doJK1 |
perform first order jackknife correction? Automatically set to |
doJK2 |
perform second order jackknife correction? Caution: this can take a very long time to run. Default value FALSE |
jkscorethresh |
square L2 norm threshold for leave-one-out and leave-two-out estimates to be included in the computation of the first and second order jackknife corrected estimate, respectively |
jkvalrange |
range of leave-one-out and leave-two-out beta/xi estimates to be allowed in the computation of the first and second order jackknife corrected estimate, respectively |
This is the main function of the package, used for fitting the linear ERR model in matched case-control data. Use this function to estimate the MLE (including a profile likelihood confidence interval for the dose effect) and to perform first and second order jackknife corrections.
The model being fit is HR=∑(1+β d_l)exp(α_l+X^Tγ), where the sum is over organ locations. Here β is the dose effect, α are the location effects and γ are other covariate effects. The model can be reparametrized to HR=∑(1+exp(ξ) d_l)exp(α_l+X^Tγ) using repar=TRUE
. In the original parametrization, β is constrained such that HR cannot be negative. There are different choices for the design used to estimate the parameters: mean organ dose, CCML, CL, and CCAL. Mean organ dose (ccmethod='meandose'
) uses the mean of the supplied location doses and compares that mean dose between case and matched controls. The other choices (CCML, CL and CCAL) use the tumor location for the case and compare either only between patients (CCML), only within patients (CL) or both between and within patients (CCAL). CCML only compares the same location between patients, and hence cannot be used to estimate location effects. Similarly, CL compares within patients and cannot be used to estimate covariate effects other than dose, meaning corrvars
should not be supplied for CL.
For one-dimensional models (i.e., mean dose or CCML without additional covariates), the Brent algorithm is used with a search interval (-10,log(uplimBeta
)) when repar=TRUE
and (L,uplimBeta
) otherwise, where L is determined by the positivity constraint for HR. For other optimizations, the L-BFGS-B algorithm (with constraint uplimBeta
) is used when repar=FALSE
, and the unconstrained Nelder-Mead is used when repar=TRUE
. For details refer to the function optim
, also for fitopt
settings. Note that when supplying ndeps
to fitopt
, a value needs to be specified for every free parameter in the model. For more flexibility in optimizion, use linERRloglik
and optimize directly.
The jackknife procedure allows for filtering of the leave-one-out and leave-two-out estimates, which is important as the model can be unstable and produce extreme estimates. All estimates reaching the maximum number of iterations are excluded, as well as estimates larger than uplimBeta (if applicable). Further, the user can set a threshold for the square L2 norm of the score for an estimate (default .01), as well as an allowed value range for the β/ξ estimate itself. When the jackknife is run, the output object contains an element details
, allowing the user to inspect the produced leave-one-out and leave-two-out estimates.
Object with components MLE
and jackknife
. MLE
has components:
coef |
estimated model coefficients |
sd |
estimated standard deviation for all coefficient estimates |
vcov |
variance-covariance matrix for all estimates |
score |
score in the MLE |
convergence |
convergence code produced by the optimizer (for details refer to |
message |
convergence message produced by the optimizer |
dosepval |
p-value for the LRT comparing the produced model with a model without dose effect. Note that the null model this is based on uses the same optimization algorithm used for the MLE, meaning one-dimensional Nelder-Mead is used when |
profCI |
the 95% profile likelihood confidence interval. In some cases one or both of the bounds of the CI cannot be obtained automatically. In that case, it is possible to use the |
fitobj |
Fit object produced by linearERRfit |
jackknife
has components firstorder
and secondorder
. Both of these have components:
coef |
the jackknife-corrected coefficient estimates |
details |
data frame with information on leave-one-out or leave-two-out estimates, with columns:
|
Note that the details
for the second order jackknife only include leave-two-out estimates. To access leave-one-out estimates, use details
for the first order jackknife.
1 2 3 4 5 6 7 | data(linearERRdata1)
fitCCML <- linearERR(data=linearERRdata1, set=1, doses=2:6, status=8,
loc=7, corrvars=9, repar=FALSE, ccmethod="CCML", doJK1=TRUE)
fitCCML$MLE$coef
fitCCML$jackknife$firstorder$coef
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.