Control Parameters for logistf Profile Likelihood Confidence Interval Estimation

Sets parameters for modified Newton-Raphson iteration for finding profile likelihood confidence intervals in Firth's penalized likelihood logistic regression

logistpl.control(
  maxit = 100,
  maxhs = 0,
  maxstep = 5,
  lconv = 1e-05,
  xconv = 1e-05,
  ortho = FALSE,
  pr = FALSE
)

Arguments

maxit: The maximum number of iterations
maxhs: The maximum number of step-halvings in one iteration. The increment of the beta vector within one iteration is divided by 2 if the new beta leads to a decrease in log likelihood.
maxstep: Specifies the maximum step size in the beta vector within one iteration. Set to -1 for infinite stepsize.
lconv: Specifies the convergence criterion for the log likelihood.
xconv: Specifies the convergence criterion for the parameter estimates.
ortho: Requests orthogonalization of variable for which confidence intervals are computed with respect to other covariates
pr: Request rotation of the matrix spanned by the covariates

Value

maxit: The maximum number of iterations
maxhs: The maximum number of step-halvings in one iteration. The increment of the beta vector within one iteration is divided by 2 if the new beta leads to a decrease in log likelihood.
maxstep: Specifies the maximum step size in the beta vector within one iteration.
lconv: Specifies the convergence criterion for the log likelihood.
xconv: Specifies the convergence criterion for the parameter estimates.
ortho: specifies if orthogonalization is requested.
pr: specifies if rotation is requested

Details

logistpl.control() is used by logistf to set control parameters to default values when computing profile likelihood confidence intervals. Different values can be specified, e. g., by logistf(..., control= logistf.control(maxstep=1)).

Author

Georg Heinze

Examples

data(sexagg)
fit2<-logistf(case ~ age+oc+vic+vicl+vis+dia, data=sexagg, weights=COUNT, 
    plcontrol=logistpl.control(maxstep=1))
summary(fit2)
#> logistf(formula = case ~ age + oc + vic + vicl + vis + dia, data = sexagg, 
#>     plcontrol = logistpl.control(maxstep = 1), weights = COUNT)
#> 
#> Model fitted by Penalized ML
#> Coefficients:
#>                    coef  se(coef) lower 0.95  upper 0.95       Chisq
#> (Intercept)  0.12025405 0.4763429 -0.8185591  1.07315122  0.06286298
#> age         -1.10598131 0.4149021 -1.9737884 -0.30742514  7.50773092
#> oc          -0.06881673 0.4344026 -0.9414363  0.78920202  0.02467044
#> vic          2.26887464 0.5384872  1.2730216  3.43543273 22.93139022
#> vicl        -2.11140817 0.5320395 -3.2608611 -1.11773495 19.10407252
#> vis         -0.78831694 0.4089620 -1.6080879  0.01518468  3.69740975
#> dia          3.09601166 1.5052197  0.7745682  8.03029352  7.89693139
#>                        p method
#> (Intercept) 8.020268e-01      2
#> age         6.143472e-03      2
#> oc          8.751911e-01      2
#> vic         1.678877e-06      2
#> vicl        1.237805e-05      2
#> vis         5.449701e-02      2
#> dia         4.951873e-03      2
#> 
#> Method: 1-Wald, 2-Profile penalized log-likelihood, 3-None
#> 
#> Likelihood ratio test=49.09064 on 6 df, p=7.15089e-09, n=239
#> Wald test = 31.96835 on 6 df, p = 1.654713e-05